Thursday, December 25, 2008

The Case of the Missing FX Deposits

The title sounds like a Sherlock Holmes novel (or a development economics paper), but is in fact from a recent speech by Ersin Ozince, the CEO of Is Bankasi (and the head of Banks Association of Turkey). First, it was thought that he was talking of of total deposits, but he clarified the next day, saying that the 10% was referring to FX deposits. Mr. Ozince does not mention which dates he was referring to (or at least, it was not reported), so using the latest CBT weekly bulletin, I graphed FX and TRY deposits for the last 13 weeks (ending with early December, figures in thousand TRY). For added entertainment, I also graphed the weighted exchange rate (50% EUR, 50% USD) on the secondary axis (RHS).

Just looking at this graph brings out quite a few interesting points:

It is just a matter of algebra to use weekly the exchange rate figures to verify Mr. Ozince's claim. By using this particular weighted exchange rate, I am assuming that deposits are equally divided between EUR and USD, which gets me close to the 10%figure Mr. Ozince was mentioning.

But this above calculation would only be rough estimation because of what a retail banker friend of mine has aptly called "the parity illusion". In this time period, the dollar gained quite a bit against the euro- this was reversed in December in the wake of the end of a great wave of deleveraging and the Fed's strong policy actions. Once you clean the data of exchange rate movements and the parity illusion, you'd probably find that the total deposits have remained stable, with the increase in TRY deposits compensating for the decline in FX deposits. So the case of the missing FX deposits is really a case of misplaced deposits.

But to undertake this exercise, you'd need to know the share of EUR and USD deposits in the first place. Theoretically, it should be possible to estimate this by using EURUSD, EURTRY and USDTRY. I have to think through this, but concentrating on an extended time period when one of the latter two was constant might do the trick.

The same idea could be reversed: It could also be possible to model the depositors' switching between TRY and FX deposits (and in between FX deposits, to add a further layer of complication). The main issue is that both quantities (deposits) and prices (USDTRY, EURTRY) would be endogenous, so you'd want to do a VAR perhaps, but I haven't done any literature review yet. I am just thinking aloud for now, but all of this would make a couple of great papers for the time series class I am teaching at Izmir Economics University.

No comments: