The author has been described by News Ltd as an "iconoclast", "Svengali", a pollie's "economist muse", and "pungently accurate". Fairfax says he is a "Renaissance man" and "one of Australia’s most respected analysts." Stephen Koukoulas concludes that he is "85% right", and "would make a great Opposition leader." Terry McCrann claims the author thinks "‘nuance’ is a trendy village in the south of France", but can be "scintillating" when he thinks "clearly". The ACTU reckons he’s "an enigma wrapped in a Bloomberg terminal, wrapped in some apparently well-honed abs."

Sunday, November 8, 2009

If I ran APRA (updated)...

One of the problems with managing the Byzantine nexus between asset price and credit cycles, and the ramifications of such for the real-economy, is that policymakers have historically had very poor credit data. As the RBA spent considerable time highlighting in its latest Financial Stability Review, the measurement of seemingly simple statistics like mortgage default rates is an exceedingly complicated exercise both within countries and across nations. While this might sound esoteric to some, getting an understanding of the risks accompanying the $1 trillion worth of outstanding Australian mortgage debt is vital to all of our welfare.

Yet even default rates are an ex post expression of financial duress. That is to say, you only find out after the event. In fact, the probability of default tends to peak a full two to three years after the date on which a loan is originated. A more valuable leading indicator of financial duress that could be used to predict changes in default rates over time would be live data on the quality of the assessment standards employed by lenders when they extend credit.

To the best of my knowledge, it has been awfully difficult for policymakers to get access to this critical real-time information. This is certainly true in the residential mortgage-backed securitisation sector, where issuers do not ordinarily provide investors with data on the debt serviceability ratios underpinning the individual loans that they are vending into the market.

Getting data on dynamic changes in lending standards over time is crucial precisely because these standards are not static. History tells us that the severity of credit standards is highly pro-cyclical—money is easy to access during the good times, and hoarded by suddenly risk-averse bankers during the bad.

In many ways, the GFC was simply an echo of the coincident asset price and credit boom (and subsequent bust) that occurred during the late 1980s with the distinguishing characteristic that twenty years’ hence global capital markets are much more interdependent due to the profound information and communications technology revolutions that have literally changed our way of life (aka the Internet and associated innovations). Oh and substitute in sub-prime for junk bonds.

As Lindsay Tanner thoughtfully observed in one of the Melbourne Institute panel sessions last week, another important difference this time around is that consumer, business and institutional investor confidence appears to be even more fickle and volatile than in days gone by as a function of the mind-boggling speed with which information in transmitted around the globe.

So when in the 1980s households had to wait for information to slowly percolate its way through the print and television news cycles, today consumers are fed this content in shockingly voyeuristic fashion. It’s like capitalism has become a perverse version of reality TV. And the velocity of this process means that the editorial and diligence standards that were previously applied to interrogating information have been heavily diluted (or, at the very least, consumers absorb much more raw content than they have ever before).

The tyranny of distance is long gone—today we face the tyranny of virtual proximity. And so we had the butterfly effect that was US sub-prime borrowers defaulting on their loans quickly causing chaos around the world, including, but not limited to, the first run on a UK bank since 1866, the nationalisation of many private banking systems, and wholesale changes of government.

These new dependencies between economic events, information, consumer and business sentiment, and the latter's feedback into actions again warrant detailed study by academic researchers. They also mean that policymakers and politicians have higher duties of care to faithfully communicate with the nation as opposed to seeking to mercenarily exploit these relationships for short-term electoral gain (eg, by exaggerating the nature of problems we face).

All of this is a rather round-about way of saying that regulators also need to think more creatively about how they monitor, measure and ultimately manage risk. And in this context, I have a policy idea.

If I ran APRA, or perhaps the RBA’s Financial Stability Department, I would establish a central electronic clearinghouse of all residential and business credit originated in Australia. For simplicity’s sake, let’s call it the National Electronic Credit Register (NECR).

If you think about it, credit is effectively an over-the-counter (OTC) contract. There is no centralised exchange novating the relationship between the parties as we see, for instance, with securities listed on the ASX. And as we discovered during the GFC, one of the profound shortcomings associated with OTC markets is that they effectively eviscerate transparency. The only people who know what is going on are the counterparties.

Now in Australia, APRA and the RBA collect a great deal of ex post facto credit data. But this is normally aggregated information and does not necessarily tell them anything about the individual loan-by-loan risks. It also does not necessarily furnish them with any insights about the ex ante, or before the event, credit assessment parameters employed by lenders.

The establishment of NECR would presumably be very straightforward. All Australian lenders have electronic lodgment processes and there are standardised communications formats that allow lenders to communicate with one another (in the mortgage market this is known as LIXI (or the “language of lending”)).

APRA, the RBA, and ASIC (to cover non-banks) could, therefore, simply insist that any licenced entity involved in the creation of residential and business credit sends NECR a simple little data packet upon the settlement and, notably, discharge (ie, repayment) of every single loan. The lender’s transmission to NECR would contain:

1) A unique loan identification code (so that NECR can track the loan);
2) The loan amount;
3) The loan type (eg, 3 year fixed)
4) The interest rate;
5) The settlement/discharge date;
6) The collateral value (eg, property value);
7) The collateral address; and
8) A nationally-defined debt serviceability standard measuring the ability of the borrower to meet the repayments on the loan (all lenders use these in one form or another, so it should be easy to define a standard metric that they have to supply).

Libertarians could be appeased by noting that it would not be necessary to provide any identifiable borrower details (ie, their name) absent the street address associated with the collateral asset.

NECR would revolutionise the RBA and APRA’s approach to risk-management. Allow me to illustrate a few examples. Regulators worry about system-wide debt levels and the rate of change in credit over time. But what gets them worked up even more is the distribution of that risk. That is, those borrowers sitting in the tails that carry the highest hazards. But disaggregating this information when banks are sending you summary statistics is clearly an arduous task (obviously the regulators insist on some disaggregated data).

So rather than simply calculating a system-wide loan-to-value ratio (LVR) by dividing the total amount of mortgage debt (circa $1 trillion) by the total amount of residential property outstanding (around $4 trillion), or for the pedants, the amount of property with mortgage debt held against it (about half based on the 2006 Census), the regulators would be able to measure the individual LVRs of every single loan in the country. And they would get live updates on changes in those LVRs as loans were regularly refinanced, which they are.

Since the average home loan’s life is only around 4 years (due to refinancing), it would not be long before they had individual records on all outstanding debt.

Most significantly though, real-time data on the specific lending parameters employed by institutions (as proxied by the debt serviceability metric required by government) would allow authorities to act assertively in the upswing of concurrent asset price and credit booms rather than picking up the pieces after the bubble has burst.

There is also technology now available that enables lenders to revalue every single home in Australia on a daily, weekly or monthly basis. These Automated Property Valuation Models (AVMs)* are highly accurate on a “portfolio” basis. Armed with this new weaponry, regulators could literally quantify the LVRs and equity held in every home in the country as frequently as they desired in order to compute the fallout associated with two, three, four or five standard deviation property price falls (or, put more broadly, fat-tail events). Now that’s what I call 21st century policymaking. Sadly it will doubtless take a long-time before anyone picks it up...

Update: My old buddy, Joshua Gans, has had a similar idea for a different purpose. Joshua tells me he would like to see a NECR style facility to enable (a) more efficient switching of customer accounts between banks and (b) to minimise the information asymmetry problems that plagued the CDOs and SIVs in the US wherein policymakers had difficulties actually identifying which loans were in which portfolios, and efficiently negotiating with borrowers who are in distress. A centralised database of all customer accounts would potentially address both of these problems.

Christopher Joye writes a blog for Business Spectator. His “after-hours” blog can be found here.

*Several companies provide AVMs locally, including Fist American Core Logic and RP Data-Rismark.