I’ve been wondering about what the UK housing market crash is going to look like, and how far prices might fall.
A natural point of comparison for the forthcoming crash is the bottom of the last crash. Looking at the raw numbers from the Nationwide seasonally adjusted representational house price series, the bottom of the last crash was in 1992-11, when the representational house price was £49,602. The following graph shows the current bubble (the red line) relative to that minimum (the green line):
Of course, it’s misleading to compare house prices from 1992 with house prices today, because incomes have increased. So I ought to scale the old prices forward using a measure of average earnings. This is because the cost of renting varies roughly in proportion with earnings, and because a plausible way in which the crash comes to an end is that the cost of buying drops below the cost of renting, which means that people who are buying for value enter the market. Some of these are renters looking to save money by buying, and some are investors looking to buy-to-let for cashflow, as distinct from the speculators buying-to-let for capital gain who have pushed the bubble to such heights.
The following graph uses the National Statistics seasonally adjusted average earnings index to scale house prices relative to the bottom of the last crash (the green line). Note that according to this measure of affordability, the bottom of the last crash was in 1996-01, substantially later than the minimum price in 1992.
But this isn’t the whole story, because most people in the UK borrow money at a variable rate based on the Bank of England base rate (or at fixed rates but only for very short terms) so I really ought to take into account interest rates as well. Buyers for value enter the market when rental yields are greater than the cost of borrowing: rental yields go up with earnings, as discussed above, but the cost of borrowing depends on interest rates.
In the following graph I’ve used the Bank of England monthly average of the official base rate (the blue line) to further scale the bottom of the last crash by the cost of borrowing as well as by average earnings (the green line). I’ve applied the scaling quite naively, just dividing the price by the interest rate. According to this measure of affordability, the bottom of the last crash was in 1994-06.
This graph is surprising to me because it shows prices tracking affordability quite closely until 2002 or 2003, when the bubble really takes off. If you had asked me to guess, I would have placed the start of the bubble substantially earlier, in 1998 or so. (But the timing of house price movements does vary from place to place, and Cambridge was among the earliest places to experience a bubble.)
From the last graph, it is clear that if prices fall to the same level of affordability relative to earnings and interest rates as they did in the mid-1990s, we are looking at around a 50% fall, back to levels last seen in 2002.
How fast might prices fall? Housing crashes tend to be quite slow (compared to stocks) because houses are not liquid assets. As long as sellers are not compelled by penury to sell or default, they can continue stubbornly asking for unrealistic prices long after the supply of buyers at those prices has completely dried up. So the crash initially manifests itself as a reduction in the volume of sales rather than a drop in asking prices. Only when enough people are compelled to sell, whatever the loss, that no-one can continue to pretend that things have not changed, does the fall in prices because apparent.
An important difference between this bubble and the late-1980s bubble is that this time there are a lot more buy-to-let speculators. These people are highly leveraged (their debts are high relative to their assets) so small changes in asset prices make a big change to their situation. As it becomes more difficult to borrow money, many of these people will find that they cannot refinance at affordable rates when their teaser rates come to an end. This means that they must sell their properties or the bank will repossess. They won’t be in the position of the homeowner with negative equity but an affordable mortgage, who can afford to hang on as long as he doesn’t move house. The speculators will have to lower their price to sell.
So the crash might be quite rapid once it gets going.
Update . I couldn’t find a data series for median household income, so I’ve cobbled one together out of what feels like bits of string and glue.
National Statistics publishes an annual summary report, “Effects of taxes and benefits on household income”, and editions since 1997 are available on the web. Buried in the appendices are tables of household incomes by decile, from which the median household income can be extracted. This gives an annual series from 1997 to 2005, shown here:
|Financial year||UK median household income|
I’ve treated these as spot figures for October of each year and interpolated linearly between the annual figures to get a monthly series. Prior to 1997 and after 2005 I’ve used the average earnings index (AEI) to scale the figures. (This is very wrong, since surely one of the factors in the bubble has been the increasing number of two-income families, but this won’t show up in the AEI. If you have a better suggestion, let me know.)
The following graph shows the representational house price divided by the median household income (red line), and the payments on an interest-only loan for the representational price at the Bank of England official bank rate plus 1.5%, divided by the median household income (green line).
Nick was spot-on when he noted in the comments that according to this measure, mortgages are more affordable now than they were in 1991.
I extended the analysis back to 1975, which is as far back as the Bank of England base rate series goes. I scaled the household income backwards using the average earnings index, which is misleading for the reason noted above. The Nationwide representational price series is quarterly prior to 1991, so I interpolated linearly to get a monthly series. Here’s the result:
The annual periodic component that you can see prior to 1990 is due to the seasonal component of the AEI—the ONS removed this from the post-1990 data series.
We should be cautious in extracting lessons from this graph. The series attempts to compare median household income with a representational house price. However, the rate of house ownership increased from about 50% to about 70% over this period (most of the increase being in the 1980s). So the median household is not a good proxy for the typical home-owning household over the long term.
Also, you shouldn’t take the actual numbers seriously either. It’s clear that the median household is not buying the representative house!
One more graph to show how misleading my figures are. This compares my series (for repayments on an interest-only loan for the representational house price as a proportion of the median household income), with two ONS series for the average repayments on a first-time (resp. subsequent) mortgage as a proportion of average household income.