What determines bank ratings? Management of Financial Institutions

The Logit Model

2. 
   Variables Description
   

We selected 12 variables relating to the profitability, capital, liquidity, asset quality and the growth in order to see if those variables can explain the ratings, as shown in the following table.

1. 
   Profitability
   

As you can suppose, a higher profitability will lead to a better rating of the concerned financial institution. We expect thus the variables of profitability below to be positively correlated with the variable “rating”.

• 
  ROE:
  

• 
  Non interest income/ Revenues:
  

• 
  Non interest expenses/ Revenues:
  

2. 
   Liquidity
   

• 
  Liquid assets/ total deposits & borrowings:
  

• 
  Liquid assets/ total assets:
  

3. 
   Capital
   

• 
  Equity/ Net loans:
  

• 
  Common equity/ Net loans:
  

4. 
   Asset Quality
   

Default risk gives a lower rating and has an important influence on the rating especially in commercial and saving banks. Good proxies to measure that risk are:

• 
  Loan loss provisions/ Total assets: As already explained this measure seems us more a measure for credit risk than for profitability. The reason to include it is obvious. The higher this ratio the more it is exposed to the risks that the illiquid loans have.
  

• 
  Loan loss provisions/ Total loans(expected): We look at the influence of the provision make on the expected total loans. The higher this ratio the more defaults they expect and it is thus a negative relation to the rating.
  

5. 
   Growth
   

Before explaining the methodology of our paper we looked at different variables in different categories (like liquidity, capital, asset quality and profitability).

After doing a PCA analyses to identify the most important variables as stated in the previous paragraph we include the variables in a factor analyses done by SPSS. SPSS identified four factors that we will use for the logit model later on to see the impulse on the rating. Before looking at the factors within the model we check if there are any correlations between the variables which can distort our results.
As you can see there are a lot of correlations. This would mean to lose variables that are highly correlated. We tried to look for variables that aren’t correlated to replace the one we have to loose. After founding that all variables where uncorrelated we could do a factor analyses to see which variables determined which factor. We used SPSS who decided like stated above that 4 factors are the most appropriate.

As stated we have 11 variables after replacing or deleting the correlated ones that explained following factors.