Monday, 13 February 2017

Bubble-trouble with Australian/NZ mortgage risk-weightings

Approaching a decade on from the global financial crisis (GFC), I continue to remain amazed by how little the world has learnt. Indeed, Australia and New Zealand, for instance, remain in the grips of record property bubbles at present and are repeating many, if not most, of the same mistakes. The same can likely be said of Canada, and perhaps the UK as well (although I'm less confident on the latter).

While all sorts of new banking regulation has been proposed and implemented in the crisis' wake, the root cause of the crisis does not appear to have been either recognized or addressed. Consequently, the same fundamental mistakes are recurring, but merely in a different guise - namely the use of artificially-low mortgage risk weightings. But first some quick background.

The true cause of the GFC

In my opinion, the fundamental root cause of the GFC was a widely-held assumption by banks and other financial market participants that past, statistically-based measures of home mortgage default rates could be used as a reliable indicator of future default risk.

At face value, it seemed to be a reasonable presumption. After all, the past 70 years of data indicated that individual home mortgage default rates in any given year were low. Furthermore, in the event of default, there was typically sound collateral coverage, and so recovery rates in the event of default were typically high. That could be expected to continue, so the thinking went - after all, home prices almost always rose over time. 70 years of data proved it.

This assumption allowed vast bulks of mortgages to be securitized, rated, and sold around the world with limited scrutiny, because everyone knew these mortgages were as 'safe as houses'. When this assumption was called into question, the flow of credit was interrupted and house prices promptly crashed, because the availability and cost of credit is what determines the level of house prices.

The core problem with these models was that they considered mortgage default risk in much the same way as an auto insurer might look at their risk exposure from car accidents. In other words, they assumed individual defaults were uncorrelated and independent risk factors. Sure, on any given day an individual borrower could lose their job and default on their mortgage, in much the same what that any individual could crash their car. But these risks are diversifiable, and across a large portfolio, are low risk on a system basis, with highly predictable default probabilities.

The flaw with these models, to extend the insurance analogy, is that they didn't consider 'catastrophe risk' - e.g. the idea that a 100-year storm could sweep through town and damage a significant portion of the auto fleet in one fell swoop. Insurance companies do consider these risks, but mortgage default models did not. That was fine for 70 years (long enough for the Great Depression to be forgotten). But when a nationwide housing crash occurred, the risks that were assumed to be independent suddenly became highly correlated - something the models had not taken into account. The result was financial catastrophe.

Furthermore, these models not only failed to take into account system 'catastrophe risk', but were also active contributors to the formation of the bubble in the first place. This should have been predictable to thoughtful observers (and was to many - see The Big Short).

With interest rates very low during 2001-05 after the bursting of the tech bubble, investors needed yield. So they started to buy higher-yielding mortgage securities, which were assumed to be low risk (we all know what assumption is the mother of). Banks did what banks do - as intermediaries, seeing a large and growing demand for higher-yielding mortgage securities, they rushed to manufacture more of them to meet that demand. They needed raw material - mortgages - so they paid mortgage brokers to source them more loans. Mortgage brokers lent to anyone they could to meet the demand from Wall Street. Lending standards fell and the result was that anyone could get a loan, which drove the bubble. When the buyers of those mortgage securities suddenly realized they were taking more risk than they realized, they stopped buying them, and the whole edifice crumbled. This should have been predictable to regulators, and the crisis would never have happened if dynamic risk-models actively took into account systemic risk factors. 

The same mistake is being made with low mortgage risk-weightings 

Unfortunately, little has changed, and the problem is now manifesting in the way bank regulators determine mortgage risk-weightings, which much like the above, seem custom-designed to both overlook systemic risk while at the same time maximizing the existence of that risk. Central banks have also been slashing interest rates again. When I look at Australia and NZ, it almost appears to me as if bank regulators, politicans, and central bankers looked at what happened in the US during 2002-09 and said "hey, that seems like a good idea; let's try it".

In the banking industry, risk-weightings are used to determine how much capital a bank needs to hold against a given block of risks it is taking. For loans that are considered riskier than average, a risk weighting above 1x will be applied, which requires proportionately more capital be held against such loans. Conversely, loans or assets that are considered lower risk require a less than proportionate amount of precautionary capital be held.

What is extraordinary is that the risk-weighting applied to home mortgages in Australia is only 16% (although APRA is considering raising this to 25%). Bank regulators deem home mortgages to be so safe, that only about 1/6th of the capital backing a typical loan need be held for home mortgages. Put another way, it means banks can make six times as many mortgage loans for every dollar of capital held vs. assets with a par risk weighting.

This is pure folly for a number of reasons:

*Firstly, it makes the exact same error as the mortgage securitization models in the US made pre GFC - namely looking at mortgage risk from a purely uncorrelated, individual basis, using historical data as a guide, instead of via a thoughtful consideration of systemic risk factors. 

*Secondly, low-risk weightings have contributed - alongside low interest rates - to the development of a housing bubble in Australia/NZ in the same way false securitization model appraisals led to an excessively loose flow of credit into the mortgage market in the US pre-boom, because they have encouraged banks to lend more to the housing sector in order to maximise their returns on equity (and management bonuses). Indeed, about two-thirds of Australian banks balance sheets are exposed to home mortgages - a very high amount.

*Lastly, no adjustment is made to risk-weightings to reflect the riskiness of the underlying collateral at different points in the cycle (APRA's recent proposal is too little too late). After a housing bubble has popped and house prices are low (and capitalization rates high), housing loans are not very risky. However, when house prices are very high (and capitalization rates very low), they are much riskier (see the convexity discussion below). Constant risk-weightings assume that risks are equal at all points of the cycle, but this is patently false. This is an inexcusable regulatory oversight and failure of leadership in my view - particularly given that the GFC should have alerted regulators to these risks.

The banks have also exercised almost no self-restraint, and have behaved in exactly the way main street loathes - i.e. taking huge risks to maximize short term profits and bonuses.

The world, unfortunately, has learnt nothing.

Convexity maximizes the risks at this point in the cycle

It is important to appreciate that the convexity characteristics of property prices and other yield assets become fairly extreme at low interest rates/capitalisation rates. For a property with a gross yield of 10.0%, a 100bp increase in interest rates and corresponding capitalization rates would only reduce the property's value by about 10%. Furthermore, in an environment of 10% interest rates, you probably have at least 3-4% inflation, and so over the course of 2-3 yrs, rents would rise by about 10% and the adjustment would be a wash. Property prices would merely stabilize for a period. Something akin to this has been the typical experience historically.

However, at a gross yield of 2.5%, a 100bp increase in interest rates and capitalization rates would reduce property values by about 30% (merely repricing gross rental yields to 3.5%). This is convexity - fancy bond investor parlance for the fact that capital values change much more significantly in response to a given basis point change in interest rates when yields are low vs. when they are high.

This is important, because interest rates have not been this low before historically. Consequently, we have no reliable data on how property prices will perform in a future period where interest rates rise from very low levels. If the US experience during 2005-08 is any indication, however, it could be pretty ugly. Needless to say, the risk of a 80-90% LTV mortgage is much higher when interest rates and cap rates are low than when they are high, due to these convexity characteristics, and yet the regulatory risk-weighting is the same.

A quite shocking article was published recently in Australia that indicated 20% of Australian households felt they would struggle to meet ongoing mortgage payments if interest rates were to rise by merely 50bp. 50bp off record lows! That is a truly terrifying statistic, and if it is true, the application of a 16% risk-weighting to mortgages is folly in the extreme.


As should be apparent from the above, I am steering well clear of the Australian banks with my investment activities on the long side. I actually think there is a not-inconsiderable risk they prove to be insolvent and end up needing to be bailed out by the Australian tax-payer, as a lot of their putative 'profits' rely on perpetually rising house prices and low interest rates to be sustainable, and in the meantime, they are paying out high fractions of their paper earnings in dividends. I also own no physical property in Australia and NZ and have absolutely no desire to do so.

I have looked at ways to try to short the banks but to no avail. The ords are expensive to short and timing is extremely hard to get right. Leverage is limited when shorting the ords as well - the most you can make is 100% of your initial at-risk capital, prior to carry costs (which are high). In addition, a housing crash and banking crisis in Australia is only one of many potential outcomes. The existence of systemic risks does not guaranty that a crisis will happen - only that the potential for one exists. A number of factors have to come together for the powder keg to actually explode (a sharp and sustained rise in international wholesale interest rates could do it). I am not predicting a crash - only pointing out that the risks of one occurring a real.

Ideally, you want to own derivatives with large asymmetric payoff potential if a blow-up does occur. Paulson, for instance, bought securities pre-GFC where the downside was 2-3% of his portfolio but where the maximum payoff potential was a 10x gain on his entire portfolio. Needless, to say, he did extremely well in 2008. Unfortunately, as a small-time punter I don't have the resources to walk into an investment bank and ask them to structure me custom derivatives that would allow me to profit from any A/NZ housing bust, in the manner Burry and Paulson et al did in the US pre 2007. CBA CDSs seem cheap, but the Australian government will likely bail out the Australian banks, wiping out shareholders but protecting debt-holders. Buying long dated puts on the Ausy dollar could be one way to play it.

If anyone has any idea I'm all ears.