Friday, January 16, 2015

What's in a name? Bubbles.

This is more in a back and forth between this blog and Idiosyncratic Whisk.

First we have to discuss definitions of a bubble before we can move on to talk about if housing in the 90s through aughts qualifies.  Most people take for granted that housing was a bubble so I appreciated that there are (apparently quite bright given his blog) dissenters like Kevin questioning some huge assumptions.  For my money something is a bubble not because it deviates from historical norms, or because its price increases are "to large"-  I think what makes a bubbles is the fact (or threat) that a change in the sign of growth (not only from positive to negative, but also from positive to zero, or from zero to negative) causes a feedback loop that pushes prices down.  Ponzi schemes are classic bubbles- once new investors cannot be corralled into keeping payouts high the whole thing disappears as one person pulling out their money causes lower returns which causes others to pull out their money.  The fact that a person cannot pull their money out of Social Security (at least not easily) prevents it from this Ponzi like collapse, even though the other fundamental aspects are the same. 

I will summarize Kevin's position here- he has written a ton about it so I won't do it justice in nuance but the general gist is that with very low nominal rates of interest it makes sense to borrow large amounts (to lever up) in exchange for the cash flow (in this case the implied rent).  This is similar- but definitely not identical- to the idea that was floated during the boom that house prices were being supported by lower interest rates because lower rates meant lower costs of purchase and so a person could purchase "more house" (ie a higher prices house) for the same monthly payment.  Kevin also provides a timely quote in the comments of one of his recent posts

I prefer to begin analysis of homeownership by thinking of home values as an all-cash investment. There are ways that mortgages might effect market prices, but those are tweaks from the all-cash intrinsic value.

 This is a major point of disagreement for me.  By treating home ownership as an all-cash investment he eliminates the roles of banks, and banks and homeowners have very different risk profiles, and it is this risk profile where his analysis falls short.

Banks have two risks when they loan, interest rates and default rates.  Because they borrow short and lend long an increase in interest rates freezes the majority of their income at lower rates while increasing their operating costs.  Default risk is simply the risk that the borrower cannot pay back the loan, in which case the bank seizes the house and sells it off to recover as much of the loan as they can.  All banks use models that assume some default rate and some reduced sale price (frequently around 80% of market value) to calculate expected losses on loans.  This creates a potential double whammy if interest rates increase as banks face stiffer costs to keep their capital requirements up to snuff while also having to raise more to cover larger losses on defaults (higher interest rats -> lower prices -> lower recovery %s).

What can happen on the other side of the ledger is very interesting.  In the case of a fall in interest rates, and subsequent rise in house prices, is an apparent decrease in the risk of bank loans.  This happens because most of a bank's book (at least economy wide) is based on loans already made.  If prices increase 5% then a banks recovery rate (if it stays stable at 80% of market value) will now be 84% of the original purchase price for all previous purchases.  Lets say a bank's book is 10% loans made recently and 90% made in the past, then every price increase reduces a bank's default risk by 9 times that of the risk new loans carry at the higher price.  Every new marginal borrower can be riskier than the previous average in the bank's eyes.

Here we can describe a positive feedback look pushing prices higher in which banks appear to be acting sanely.  Higher prices mean a lower risk profile for the bank, which allows slightly riskier loans, this pulls more home buyers into the market which increases demand and should then push up prices and/or interest rates.  For a visual here is home ownership rates in the US
No matter the explanation for 30 years years banks rated between 63 and 66% of the population as acceptable risk as homeowners.  From 1995 to 2005 we go from ~64% (roughly the 30 year average) to 69%- more than 3 percentage points higher than the previous high shown.  The US had between 105-110 million households in 2005 so that 3-5% rise over the historical average means that over 10 years there were 3.1-5.5 million more homeowners than you would have predicted in 1995.  This HO rate increase roughly matches when price increases started to take off in earnest

Going back to my (quick and dirty) definition of a bubble is a feedback loop once we have a change in the sign of prices.  First I want to say that banks are not run by idiots, and the people writing models that project things like default risk and recovery rates are not idiots, but the way most models are written is based on historical data.  If default rates have hovered around 1% for 50 years, a number close to 1% is going into your model, if recovery rates are 80% over that span, and number close to 80% is going into that model.  Models like these tend to break down around historical anomalies, not just because of different dynamics at those points but because the data leading up to those points is usually incorporated into their models.  When loans were made in 2004 the price increases of 1995-2003 were going to be a natural part of estimating risk.

In my narrative we have moved the price of housing up and the quality of homeowners in terms of default risk down.  Implied in that second part is the fact that the remaining pool of potential homeowners is of lower quality than the historical average.  Also implied is that the lower quality a homeowner you are (if you have bought a house) the closer to the peak purchase price of was the timing of your purchase.  Now a change in the price signal- even just from positive to no growth- will cause a spiral.  Defaults will be at the high end of historical averages (and maybe higher) and the cost of those purchases will also be the highest of those on the books.  Recovery rates fall because the pool of available buyers is at an all time low as well.  Now banks do some combination of raising capital and slowing loan origination to fill these losses that are on the edges (and sometimes over the edge) of their estimates and we have the potential for a spiral without a recession kick starting or amplifying it.

There are still lots of missing pieces in the puzzle, but to return to the topic- if you treat a home purchase as a cash transaction you will fail to see how a bank's risk profile changes in response to price changes. 

Now you can tell a different story with the data shown- but it basically requires the claim that either a substantial portion of the population (3-5% percent of households) went from being a poor mortgage risk to a good one, or that banks were leaving billions on the table from 1965-1995 and then suddenly came to their senses. 


  1. the definition is important, but i don't know if the feedback loop definition one is a good one. i'm sure austrians have a numbers based definition that's more appropriate

    1. Price is not a good metric, fundementals can change and markets are allowed to be inefficient. The implication of the feedback loop is that the current price is (somewhat) dependant on future price increases. If you go by price based metrics it is hard to see the difference between speculation and bubble.