Tuesday, 7 June 2011


I've been meaning to write something on this subject for a while but I try not to discuss banks too much so it took a back seat. Eventually I got a little time to do so.

I have for some time had a professional interest in the impact of Basel III, which is probably best described as the new rulebook for bank capital and liquidity requirements; this is generally sold as (or derided as, depending on where one sits) the recipe for avoiding another financial crisis, or at least the main regulatory and supervisory ingredients of the recipe. Rest assured, this will not be the subject of this post. If you want to read up on this, I would suggest tucking into the highest-profile impact assessments to date (ping me if I've forgotten any biggies). Be sure to note the publication dates as the proposed shape of Basel III has changed over time. More on this here.
Now back to our story.

I was reading the IMF impact assessment recently when I noticed something particularly iffy in the Annex, which as we all know is the Top Shelf of statporn. The authors try to model the effects of new capital requirements by looking at the past behaviour of banks and their customers in twelve countries, including in this case Greece. That's not the iffy bit. That's the jackpot.

UPDATE: One reader has correctly pointed out that the sample the IMF is working on, at least in the case of Greece, is tiny. I will happily second their argument that it's silly to try to forecast Greek banks' or their customers' behaviour based on such figures; but in fairness to the authors, that was neither their intention nor the outcome of their work; their intention was among other things to look at how the dynamics of the banking sector differ in crisis countries, of which Greece was one. That said, the small Greek sample still represents almost all of the banks in the country, so it's probably best to treat these tables as indicative of the past behaviour of Greek banks.

I was however particularly struck by Table 9, page 29, which looks at how loan demand changed in response to different variables. What one would expect is that as the cost of borrowing goes up, demand would go down. Which it does, in eight out of twelve countries. In another two, the relationship is not statistically significant. But in Switzerland and Greece, the relationship is reversed (strongly in the case of Greece, and weakly in the case of Switzerland).

Now this I do not understand. This suggests that either Greek banking was mostly sub-prime in some way (i.e. based on a model of chasing poor quality borrowers) or that Greek borrowers flocked to the more expensive lenders. The first option doesn't really sound very convincing: most Greeks own property, which means they have access to collateral. And don't get me started on family bailouts. Moreover, Table 8 in the immediately previous page shows there was no relationship between interest income - to - asset ratios (essentially the rate at which assets were producing interest) and the bank's ratio of non-performing loans. In fact what relationship there was was very statistically insignificant, whereas you would expect a positive relationship in a sub-prime-focused sector.

There is however a very strong positive relationship between the banks' interest income to asset ratio and the log size of their balance sheets. Simply put, and if I understand this correctly, the larger the bank, the more interest it got out each Euro of its assets, although with diminishing returns. This suggests to me that either large Greek banks got more of their income from interest, or they could get away with charging higher interest rates. Or both.

Doesn't sound like a thriving competitive market. Or am I getting this wrong?  Help me understand banker friends.

UPDATE: One friend of this blog suggests that banking group data may be skewing the numbers here. It makes sense that more diversified groups with assets abroad will be earning at a higher rate. I should also thank them for pointing out the embarrassing mistake on the meaning of correlations with log assets. All fixed now I think.


  1. I can buy to some extent the result presented in the first column of IMF's table, but some of the individual countries results provide no information (at least to me). In the case of greece, they run a regression with 35 obs. I hope their predictions about the greek economy are based in a sample bigger than that.

  2. It is odd. Some quick thoughts - it might be possible that the reputation of the bank matters most than in other Western countries. Greeks, for some reason (lazyness or misinformation, maybe?), might favor long-term relationships with their banks, and be unwilling to shop around for a better deal. Possibly, the costs of changing your bank are significant. Possibly, they feel that bank contracts are not transparent enough, and something unexpected lurks in the small print. Maybe lower interests, in the peculiar Greek context (of precariousness, mistrust and disastrous economic and political history) sends the wrong signal (e.g., that the deposits are not safe in "cheap" banks).


Please remember that I am not notified of any comments and will not respond via comments.

Try to keep your criticism constructive and if you don't like something, do tell me how to fix it. If I use any of your suggestions, you will be duly credited.

Although I'm happy to entertain criticism of myself in the comments section, I will not tolerate hate speech. You will be given a written warning and after that I will delete further offending comments.

I will also delete any comments that are clearly randomly generated by third parties for their own promotion.

Occasionally, your comments may land in the spam box, which may cause them to appear with a slight delay as I have to approve them myself.

Thanks in advance for your kind words... and your trolling, if you are so inclined.