I had a conversation recently about if it would be a wise choice to make mortgage loans in Greece non recourse. In my view, a non recourse loan will mean that the collateral value is of major importance (since the borrower’s assets and income will not be liable for loan repayment after default) which means that the LTV ratio will be a function of the collateral volatility.

The problem is that (as was evident in previous posts about the construction sector), house prices are highly volatile with large swings in both directions. As a result, the necessary LTV will be quite low and lead to the banking system as a whole to shrink. To illustrate my point I ‘ve calculated the annualized percentage change for quarterly Greek house prices based on the historical series available from Bank of Greece (which appears to be a normally distributed process):

It is obvious that volatility is quite high, with a standard deviation of 9.9%. Taking 2σ (which covers the whole range of percentage changes) means that banks should be ready for a price change of 20%. Since mortgage loans are of very long maturities (usually decades) and the holding period of impaired loans can be years, the implied LTV ratio is quite low and changes the house market significantly (since any potential buyer will have to finance a large part of the house purchase with his own funds).

The next question is whether volatility follows a steady path or if there are major ‘regime changes’ along the way, which would make computing the actual volatility during the loan lifetime very hard. The graph below shows the population moving standard deviation for 4 periods (movstdevp() function in eviews). Volatility actually comes in waves which last quite a while and are not easily predicted:

Trying to find a pattern, one can test for a simple AR(1) model:

Moving stdev appears to be stationary (a unit root confirms that) with a long-run mean of 3.8% (close to the mean of the normally distributed process). Nevertheless, a quick look at the regression graph shows that the residuals are significant, hard to predict and seem to persist for quite some time. The latter means that if a bank were to liquidate collateral during these periods, it might face a loss if it based its LTV on the above (very simple) regression. A regime change since 2008 with possible (negative) auto-correlation in the residuals is also visible.

In general, the above simple calculations mean that a change to non recourse would change the housing market significantly, lowering available credit. This might lower house prices volatility (they are normally correlated with mortgage expansion) but at the cost of affordability.  Furthermore, since mortgages constitute a significant part of the asset side of any bank balance sheet, a lower LTV would mean either a lower deposit base or a need to cover deposits with other assets. At least in the case of Greece with significant long-run primary surpluses and small future demand for credit, such a policy might actually mean lower deposits. In any case, one should be explicit about the implications of non recourse loans, which are lower LTV and bank assets.

Note: The above are for illustration purposes, a more correct model for volatility would probably be a GARCH one.