1 Introduction
2 Literature review
2.1 Failure prediction methodologies
2.2 Financial ratios
2.3 Non-financial ratios
3 Data and sample
3.1 Variables selection
No | Dataset | Years | Lag | Failed banks | Non-failed banks |
---|---|---|---|---|---|
1 | CAMEL | 2010–2018 | 1 | 261 | 261 |
2 | 242 | 242 | |||
3 | 200 | 200 | |||
2 | CG | 2010–2018 | 1 | 99 | 99 |
2 | 101 | 101 | |||
3 | 96 | 96 | |||
3 | CAMEL + CG | 2010–2018 | 1 | 70 | 70 |
2 | 64 | 64 | |||
3 | 56 | 56 | |||
4 | ISS | 2013–2018 | 1 | 94 | 94 |
2 | 82 | 82 | |||
3 | 58 | 58 | |||
5 | CAMEL + ISS | 2013–2018 | 1 | 60 | 60 |
2 | 54 | 54 | |||
3 | 34 | 34 |
3.2 Data sampling
Small banks | Big banks | |
---|---|---|
Mean | 0.624 | 0.798 |
Variance | 0.236 | 0.162 |
F | 1.455 | |
P(F < = f) one-tail | 0.007 | |
F Critical one-tail | 1.281 |
3.3 Discriminatory power test
Variable | P-Value |
---|---|
Panel A: CAMEL ratios | |
TIEtoTA | 0.613 |
ECtoA | 0.804 |
TEtoGL | 0.143 |
PTItoE | 0.000 |
LtoE | 0.391 |
LAtoL | 0.358 |
NCOtoL | 0.959 |
TEtoTA | 0.110 |
TLtoTA | 0.895 |
NLLtoTA | 0.077 |
DtoA | 0.879 |
LPtoNCO | 0.379 |
ECofNCO | 0.001 |
AperE | 0.016 |
NIEtoTI | 0.001 |
IGR | 0.284 |
IBEItoA | 0.000 |
IEtoTE | 0.708 |
REtoTA | 0.000 |
CtoTA | 0.814 |
CtoTL | 0.814 |
NLLtoD | 0.040 |
GLtoTD | 0.037 |
Panel B: Corporate governance | |
BS | 0.777 |
BM | 0.131 |
BA | 0.369 |
GD | 0.953 |
BAA | 0.508 |
BD | 0.303 |
CD | 0.328 |
CPS | 0.001 |
CEOS | 0.929 |
ExecuS | 0.753 |
CEOO | 0.788 |
ExecuO | 0.572 |
CEOD | 0.894 |
ExecuD | 0.870 |
CEONEI | 0.673 |
ExecuNEI | 0.467 |
BStock | 0.375 |
CEOC | 0.192 |
ExecuC | 0.708 |
CA | 0.553 |
UVR | 0.098 |
InstitutO | 0.007 |
InsideO | 0.942 |
4 Methodology and results
4.1 Dataset 1: CAMEL ratios
Standardised canonical discriminant function coefficients | |||
---|---|---|---|
1 | 2 | 3 | |
PTItoE (Capital) | 0.113 | − 0.117 | 0.198 |
ECOofNCO (Asset) | − 0.005 | 0.083 | 0.340 |
AperE (Management) | 0.308 | − 0.109 | 0.108 |
NIEtoTI (Earnings) | 0.542*** | − 0.168* | 0.059 |
IBEItoA (Earnings) | − 0.220*** | 0.152*** | − 0.414*** |
REtoTA (Earnings) | − 0.437*** | 0.748*** | − 0.586*** |
NLLtoD (Liquidity) | 0.268** | 3.329** | − 2.294** |
GLtoTD (Liquidity) | 0.245** | − 3.747** | 2.731** |
Model statistics | |||
No. of failed banks | 261 | 242 | 200 |
No. of non-failed banks | 261 | 242 | 200 |
Eigenvalue | 0.078 | 0.053 | 0.066 |
Canonical Correlation | 0.268 | 0.224 | 0.248 |
Chi-square | 38.576 | 27.715 | 25.075 |
Wilk’s Lambda Sig | 0.000 | 0.002 | 0.002 |
Accuracy % | 61.1% | 61.2% | 60.3% |
Classification: % correct—failed | 58.2% | 58.3.0% | 55.5% |
Classification: % correct—non-failed | 64.0% | 64.0% | 65.0% |
4.2 Dataset 2: CG variables
Standardised canonical discriminant function coefficients | |||
---|---|---|---|
1 | 2 | 3 | |
CPS | 0.768*** | 0.740*** | 0.820*** |
UVR | 0.235* | 0.428** | 0.320* |
InstitutO | 0.552*** | 0.588*** | 0.618** |
Model statistics | |||
No. of failed banks | 99 | 101 | 96 |
No. of non-failed banks | 99 | 101 | 96 |
Eigenvalue | 0.111 | 0.136 | 0.111 |
Canonical correlation | 0.316 | 0.346 | 0.316 |
Chi-square | 20.449 | 25.241 | 19.846 |
Wilk’s Lambda Sig | 0.000 | 0.000 | 0.000 |
Accuracy % | 63.1% | 62.4% | 64.1% |
Classification: % correct—failed | 58.6% | 55.4% | 59.4% |
Classification: % correct—non-failed | 67.7% | 69.3% | 68.8% |
4.3 Dataset 3: CAMEL and CG
Standardised canonical discriminant function coefficients | |||
---|---|---|---|
1 | 2 | 3 | |
PTItoE (Capital) | − 0.238** | − 0.342 | 0.371 |
ECOofNCO (Assets) | 0.270 | − 0.297 | 0.155 |
AperE (Management) | 0.147 | − 0.120 | − 0.289 |
NIEtoTI (Earnings) | 0.322 | − 0.020 | − 0.194 |
IBEItoA (Earnings) | − 0.029 | 0.257 | − 0.755** |
REtoTA (Earnings) | − 0.262*** | − 0.285*** | − 0.405*** |
NLLtoD (Liquidity) | 2.519* | 0.228 | − 0.895 |
GLtoTD (Liquidity) | − 2.248 | − 0.155 | 1.114 |
CPS | 0.644*** | 0.578*** | 0.435** |
UVR | 0.072 | 0.355** | 0.308** |
InsitutO | 0.316** | 0.470** | 0.448 |
Model statistics | |||
No. of failed banks | 70 | 64 | 56 |
No. of non-failed banks | 70 | 64 | 56 |
Eigenvalue | 0.215 | 0.182 | 0.184 |
Canonical Correlation | 0.421 | 0.392 | 0.394 |
Chi-square | 25.791 | 20.118 | 17.662 |
Wilk’s Lambda Sig | 0.007 | 0.044 | 0.090 |
Accuracy % | 67.1% | 64.1% | 71.4% |
Classification: % correct—failed | 62.9% | 54.7% | 67.9% |
Classification: % correct—non-failed | 71.4% | 73.4% | 75.0% |
4.4 Datasets 4 and 5: CAMEL and ISS
Standardised canonical discriminant function coefficients | |||
---|---|---|---|
1 | 2 | 3 | |
PTItoE (Capital) | 0.155 | 0.004** | 0.409 |
ECOofNCO (Assets) | 0.355 | − 0.393* | 0.353 |
AperE (Management) | 0.321 | 0.337 | 0.602 |
NIEtoTI (Earnings) | 0.502 | 0.791** | 1.469 |
IBEItoA (Earnings) | − 0.196 | 0.309 | 0.503 |
REtoTA (Earnings) | − 0.564** | − 0.249** | − 0.246 |
NLLtoD (Liquidity) | 2.339* | − 1.757** | 0.631 |
GLtoTD (Liquidity) | − 1.891* | 2.572** | Excluded because of tolerance test failure |
ISSB | − 0.203 | − 0.205 | 0.579 |
ISSS | − 0.361 | 0.092 | 0.535 |
ISSC | − 0.175 | 0.209 | − 0.263 |
ISSA | − 0.164 | 0.111 | 0.101 |
Model statistics | |||
No. of failed banks | 60 | 54 | 34 |
No. of non-failed banks | 60 | 54 | 34 |
Eigenvalue | 0.118 | 0.226 | 0.388 |
Canonical correlation | 0.324 | 0.429 | 0.529 |
Chi-square | 12.455 | 20.390 | 19.830 |
Wilk’s Lambda Sig | 0.410 | 0.060 | 0.048 |
Accuracy % | 60.8% | 68.5% | 75.0% |
Classification: % correct—failed | 60.0% | 66.7% | 73.5% |
Classification: % correct—Non-failed | 61.7% | 70.4% | 76.5% |
5 Robustness test
6 Additional analysis
Standardised canonical discriminant function coefficients | |||
---|---|---|---|
1 | 2 | 3 | |
Panel A: CAMEL and CG | |||
PTItoE (Capital) | − 0.004 | − 0.361 | 0.417 |
ECOofNCO (Assets) | 0.029 | − 0.310 | − 0.015 |
AperE (Management) | 0.101 | − 0.116 | − 0.167 |
NIEtoTI (Earnings) | 0.439 | − 0.221 | − 0.077 |
IBEItoA (Earnings) | − 0.096 | 0.283 | − 0.645** |
REtoTA (Earnings) | − 0.247** | − 0.351** | − 0.400** |
NLLtoD (Liquidity) | 0.365 | 7.026 | 4.244 |
GLtoTD (Liquidity) | 0.342 | − 6.981 | − 4.077 |
CPS | 0.716*** | 0.665*** | 0.607*** |
UVR | 0.068 | 0.372* | 0.313* |
InsitutO | 0.316** | 0.196 | 0.215 |
Model statistics | |||
No. of failed banks | 52 | 48 | 38 |
No. of non-failed banks | 52 | 48 | 38 |
Eigenvalue | 0.240 | 0.225 | 0.257 |
Canonical Correlation | 0.440 | 0.429 | 0.453 |
Chi-square | 20.670 | 17.984 | 15.694 |
Wilk’s Lambda Sig | 0.024 | 0.082 | 0.053 |
Accuracy % | 67.0% | 63.5% | 71.1% |
Classification: % correct—failed | 60.8% | 52.1% | 60.5% |
Classification: % correct—Non-failed | 73.1% | 75.0% | 81.6% |
Panel B: CAMEL only | |||
PTItoE (Capital) | 0.016 | 0.618 | − 0.548 |
ECOofNCO (Assets) | − 0.145 | 0.313 | − 0.026 |
AperE (Management) | 0.257 | 0.075 | 0.247 |
NIEtoTI (Earnings) | 0.537 | 0.124 | 0.270 |
IBEItoA (Earnings) | − 0.287 | − 0.400 | 1.070 |
REtoTA (Earnings) | − 0.598** | 0.600** | 0.640 |
NLLtoD (Liquidity) | 0.546 | − 0.866 | − 3.164 |
GLtoTD (Liquidity) | 0.523 | 8.584 | 2.896 |
Model statistics | |||
No. of failed banks | 52 | 48 | 38 |
No. of non-failed banks | 52 | 48 | 38 |
Eigenvalue | 0.106 | 0.101 | 0.135 |
Canonical correlation | 0.309 | 0.302 | 0.345 |
Chi-square | 9.786 | 8.635 | 8.862 |
Wilk’s Lambda Sig | 0.181 | 0.204 | 0.254 |
Accuracy % | 64.1% | 58.3% | 55.3% |
Classification: % correct—failed | 63.5% | 52.1% | 50.0% |
Classification: % correct—Non-failed | 64.7% | 64.6% | 60.5% |
1 (%) | 2 (%) | 3 (%) | |
---|---|---|---|
Panel A: CAMEL and CG | |||
Accuracy % | 61.3 | 66.7 | 72.4 |
% Accuracy—failed | 73.7 | 75.0 | 72.2 |
% Accuracy—Non-failed | 41.7 | 54.6 | 72.7 |
Panel B: CAMEL only | |||
Accuracy % | 60.20 | 59.25 | 55.17 |
% Accuracy—failed | 65.21 | 74.8 | 61.90 |
% Accuracy—Non-failed | 50.00 | 51.20 | 37.50 |