I didn't see the loadings, but principal components are ordered with respect to the amount of variability that they explain (from the total variability of your dataset). So PC1, that is Principal Component 1, is the linear combination of your old variables that has maximum variance, or explains the most amount of the total variability. Principal Component 2 is the the linear combination of your old variables that has the next largest variance, and so on. What you may see is what variability each of your components are explaining (since they are orthogonal, what one explains, the other one doesn't). Figure out from your loadings, which ones have larger weights in PC1 and which ones have larger weights on PC2, and in that basis, choose which one is appropriate for your index.

Just a recommendation!

Cheers

I would suggest you to use the loading plot that explains maximum amount of variability. Since in 2D PCA, you need minimum of two principal components to explain the variance, use those two components that explain maximum cumulative variability. For instance if PC1 explains 50 percent, PC2 explains 15 percent, PC3 explains 25 percent, PC4 explains 10 percent. Then loading plot should be made PC1 vs PC3 since it explains 75 percent of variance. 

Thanks Jose and Asit for your valid contributions but it would have been more helpful if they are tailored in line with my actual loadings (attached excel sheet)....being my first time on PCA.

....nevertheless, thank you!

I've checked the sheet.  I can't seem to find the eigenvalues. Can you provide me the info of how much percentage of the total variance each PC explain? (If your PCA routine doesn't provide this info, grab the PC's corresponding eigenvalue, and divide it by the sum of all the eigenvalues). 

Every principle component (PC) is a linear combination of your variables. In particular, the varimax component 1 is a linear combination that is weighing only [Ret. On Equity]with a constant 0.64, [Ret. on Assets] with 0.68, [Net Int. Margin] with 0.31, and the rest aren't weighed. Since PC1 is the one that explains the  most of the total variability in the dataset, it is saying that these variables are the ones that vary the most in your dataset. Note that if you do 0.64^2+0.68^2+0.31^2 it is almost 1 (the difference is in those loadings you didn't take into account). I hope this helps to understand the results your are obtaining. 

Asit, PC's are always ordered with respect of the amount of percentage of the total variance they explain, so it is impossible that PC2 explains a larger amount that PC3.

Hi Jose, to be honest you are a patient and a good teacher. I did the estimations again and these are the eigenvalues and the percentage of total variation of each PC:

PC1 = 2.41281 / 0.4021

PC2 = 1.20436 / 0.2007

PC3 = 1.05168 / 0.1753 

PC4 = 0.636967 / 0.1062

PC5 = 0.579496 / 0.0966

PC6 = 0.114693 / 0.0191

Your explanations are quite clear on the Varimax rotation, but PC2 is also a linear combination of 3 indicators with these weights: [Net interest margin] with 0.4208, [zscore] with -0.7401 and [liq assets] with 0.5177. How do I decide between PC1 and PC2 or can I make use of both and compose 2 different indexes?

I also don't understand why you square this: 0.64^2+0.68^2+0.31^2. What's the intuition?

Sorry to bother....I am a fast learner, though!

...gracias!