More useless house price forecasts?
The FT didn’t publish my letter challenging John Kay’s article rubbishing attempts to model and forecast house prices. Last Wednesday it reported on its front page a research report by Morgan Stanley economist Professor David Miles, author of a report on housing finance for the government, headlined “Housing bust ‘likely in the next few years'”.
It is evidently based on a relatively simple model. The key variable explaining real (inflation-adjusted) house price growth, other than the expected variables of building costs and personal incomes growth, is expected future price increases. If these lose touch with reality, bubble conditions may be created. This looks like pretty useful information.
Commenting on the same report last week, FT economist Martin Wolf accepted the implications of the model. Where he, John Kay and David Miles all agree is that the timing of the change in direction and the ultimate extent of any correction – in other words the profile of the cycle through time – cannot be accurately predicted.
So? This is true of any long-term asset model, including for equities. But it does not detract from the value of such a model in quantifying mean expected returns or the longer-period bounds of probable real price movements.
This difference between the predictive value of historical data for prices or returns (particularly when inflation distortions are removed) as between short and long-period forecasts is a very important one for investors – a powerful ‘no nonsense’ distinction. Giving more importance to guessing essentially random short-term movements risks wasting information which could usefully support decisions with far greater long-run consequences.
As Martin Wolf points out, in the case of housing, unrealistic expectations about long-period returns will blind people to the cumulative damage to living standards resulting from paying even normal real interest rates over long periods. Quite right. But if the unrealistic expectations about real prices have come about indirectly because of excessive liquidity growth in economies in general and a reduction in lending standards in particular, it is likely that the next cycle will be marked by higher real interest rates, compounding the real income (or real wealth) squeeze for borrowers.
I am sticking with my ‘useless’ forecast that (national average) prices will fall in real terms by about 25% in real terms (ie a forecast with a 50% chance of being exceeded), with a low probability of a loss of about 40% (similar to the actual outturn in the early 1990s). I also reiterate my view that the more important dynamic for the greatest number of current owners is the prospect of no real growth in prices for at least the next 10 years and no more than 1% (half the historical trend) over 20 years.
I do not know whether the Miles model generates a medium-term projection but the assumption of reversion to an observed sustainable or normal growth rate in line with incomes (rejected by Kay but accepted by Miles and Wolf) is consistent with my own estimates. Models derived from historical real prices also suggest, incidentally, about half the standard deviation of equity forecasts.
Why economists and financial journalists generally pay so little attention to the nature and origins of real house price behaviour and the rationality of popular expectations is a mystery. Whatever their importance for economic prospects or policy decisions (hotly disputed amongst policy makers), house prices are clearly the main topic of conversation about money and the most popular financial topic (by far) on internet forums.