On the Likely Extent of Falls in Irish House Prices.
February 14, 2007
Looking at house price cycles across the OECD since 1970, we find a strong relationship between the size of the initial rise in price and its subsequent fall. Were this relationship to hold for Ireland, it would predict falls of real house prices of 40 to 60 per cent over a period of 8 to 9 years. House price falls tend not to have serious macroeconomic consequences, but the unusually large size of the Irish house building industry suggest that any significant house price fall that does occur could impose a difficult adjustment on the economy.
The purpose of this paper is to look at the likely behaviour of Irish house prices based on the experience of economies that have gone through similar booms. Looking at nearly 40 booms and busts in OECD economies since 1970, we find that the size of the initial boom is a strong predictor of the size and duration of the subsequent bust.
Typically, real house prices give up 70 per cent of what they gained in a boom during the bust that follows. This is a remarkably robust relationship, holding across very different OECD housing markets over more than 30 years.
Were this relationship to hold for Ireland, it would predict a fall in real house prices of around 40 to 60 per cent, over a period of 8 or 9 years. Assuming an inflation rate of 4 per cent, this would translate into an annual fall of average selling prices of around 5 per cent.
Falls of this magnitude and duration are not unprecedented internationally. For example, the real price of Dutch houses fell by 50 per cent between 1979 and 1987, while the price of houses in Britain relative to real income also fell by 50 per cent between 1948 and 1957.
Britain and the Netherlands illustrate a general point: house price falls, although uncomfortable for some households who bought around the market peak, do not usually have serious macroeconomic effects. Consumption falls as household feel less wealthy, banks experience an increase in bad loans, and fewer houses get built for a while; but the cumulative effect on growth tends not to be large or long-lasting.
Internationally, house prices boom and crash frequently, as economic theory predicts they should. Anytime an economy experiences a period of prosperity and low interest rates, house prices boom for a while, and then fall. The only unusual thing about Ireland in the past is that, thanks to the success of governments in insulating the economy from any risk of economic growth, it has never had a housing boom before.
While house price falls tend not to have serious macroeconomic consequences (the one exception is Finland at the start of the 1980s, where bad housing loans caused the banking system to collapse), they may pose some risk for the Irish economy. Typically, an industrialized economy gets around 5 per cent of its income from building new houses, around the same that it gets from household spending on recreation. Ireland currently derives nearly 4 times this amount from building and selling houses. Any sudden fall of residential investment to normal international, and national historical, levels, could have a substantial impact on national income.
What might trigger a house price collapse? The answer is that nothing is needed: price falls can occur endogenously as buyers revise their expectations of the distribution of reservation prices of other buyers. The increase in Irish housing prices has in large part been driven by expectations of future price increases: this is why house prices have doubled relative to rents since 2000.
Potential buyers can now rent for less than the interest cost of a mortgage (a million Euro house, which involves a monthly interest payment of around e4,000 can be rented for under e2,000) and need be in no hurry to buy once they no longer expect prices to keep rising. Similarly, owners of rental properties are getting returns below the rate of interest: 4% if they rent, or zero if, as in many cases, they leave the property empty (Fitz Gerald, 2005). Once they no longer expect capital gains, they should sell. The decision of potential buyers to wait and see, and for investors to sell, while supply continues to grow rapidly, will put pressure on prices. It is interesting to note that the collapse of housing booms in several US cities in the last few months was not triggered by any economic slowdown, or increase in mortgage interest rates (while US short rates have risen, long rates, on which mortgages there depend, have not), but by the feeling that houses had come to cost more than they were worth.
The rest of this paper is as follows. Section 1 rehearses the relevant economic theory of rational frenzies and wisdom after the fact in asset markets. Section 2 looks at the nearly 40 cases since 1970 where OECD economies have experienced house price rises followed by falls, and shows that the magnitude of the boom is a strong predictor of the size and duration of the subsequent bust. Section 3 shows how the stagnation of rents since 2000 while house prices doubled means that the Irish housing market has not been driven by strong fundamental demand but by a bubble. Section 4 looks at the possible magnitude and duration of house price falls, and their potential macroeconomic effects.
1 Economic theory.
The familiar efficient markets hypothesis predicts that changes in asset prices are unpredictable. The price reflects individuals’ information about asset’s present value, and changes as this information changes. Agents with good information buy, driving up the price, and those with bad information sell, driving it down.
However, instantaneous revelation of information through trade is not possible in house markets due to the very large transaction costs involved. In addition, the market lacks means for individuals to convey negative information through short sales.
As a result, housing markets are better modeled as information cascades: the actions of other agents signal their private information and can cause individuals to ignore their own signals and follow the herd (Bikchandani, Hirshleifer and Welch, 1992). Two models in the cascade literature are particularly useful for understanding the dynamics of housing markets: the rational frenzies model of Bulow and Klemperer (1994) and the wisdom after the fact model of Caplin and Leahy (1994).
Bulow and Klemperer (1994) model rational frenzies in auctions where partici- pants reveal their valuations by bidding. Suppose that there are k items available. If individual reservation prices were known with certainty, everyone would wait until the price fell to just above the reservation price of the k + 1-th highest person, and then all buy together. In practice, only the probability distribution of reservation values is known, and by bidding, or failing to bid, individuals reveal information about their valuations, allowing all participants to update their estimates about the value of the k + 1-th highest reservation price.
As a result, bidders with very different valuations have very similar willingness to pay. Price drops until one person bids. The information this reveals about the true distribution of willingness to pay can set off a bidding frenzy among the other bidders, driving up price again until it becomes clear that price is again above willingness to pay. Bidding then stops, causing prices to collapse until another bidding frenzy starts. To the extent that individuals depart from Bayesian rationality, altering reservation values in response to observed trends in prices, these effects will be amplified.
Caplin and Leahy (1994) look at investment where individuals have Gaussian signals. If the true state is bad, individuals continue to invest, driven by the domi- nating effect of past actions. Eventually, however, because signals are not bounded, a few agents get sufficiently bad signals to induce them to stop investing, causing priors rapidly to move to a belief that the state is bad, leading to a market crash and “wisdom after the fact”.
2 Mean Reversion in House Prices.
Economic theory then predicts that house prices should not follow a random walk, but should be a mean-reverting process of booms and crashes around a slowly increasing trend reflecting the growth of household income. This is what the international data show.
Large falls in real house prices in the aftermath of housing booms are common internationally. Table 1 shows the 18 cases since 1970 where OECD economies have experienced falls in real house prices of at least 20 per cent, along with the previous price rise, and the duration of the fall. It can be seen that, in contrast to stock or currency markets, falls are prolonged, usually lasting 5 to 7 years, with the Netherlands, Switzerland, and Japan all experiencing more than a decade of falls. This reflects the reluctance of sellers to cut nominal prices, meaning that inflation does most of the work in reducing real prices.
Shiller (2006) looks at three long series of real house prices: Amsterdam from 1628 to 1973, Norway from 1819 to 1989, and the United States from 1890 to 2005. In all cases he finds that although there are substantial and long lasting peaks and troughs, there is scarcely any upward long-run trend in prices.
Figure 1 shows the same pattern for smaller OECD economies: the Nordic countries, the Netherlands, and New Zealand, since 1970. The diagram shows the ratio of average house prices to disposable income but real house prices show a very similar pattern. Again, as economic theory predicts, there is considerable volatility and no sign of long-run trends. In contrast to stock price data, the tendency of prices to return to their long run average means that the size of price falls can be predicted from the size of the price rise that preceded them.
Figure 2 plots the size of increase in house prices for 17 OECD economies, against its subsequent fall.1To estimate the peaks and troughs in each series for each country, we first calculated percentage changes for each quarter. A Friedman supersmoother (implemented in the R statistics package) was then applied to the percentage changes to eliminate short-run fluctuations. Peaks and troughs were then identified as the end of runs of positive or negative changes in the smoothed series, and actual price changes calculated between these points.
Percentage rises and subsequent falls are calculated relative to different values: troughs and peaks respectively. Remember that a rise of p per cent only needs a fall of p/(1 + p) per cent to reverse it. To eliminate this complication, all rises in Figure 2 and subsequent regressions are expressed as a percentage of peak values: for example a rise from 50 to 100 is treated as a 50 per cent rise, rather than a 100 per cent one.
Figure 2 shows that there is a strong linkage between rises in real house prices and subsequent falls. There is one evident outlier corresponding to a dip in house prices in Spain that occurred in the early 1990s in an otherwise continuously up- ward trend that saw real prices quadruple between the mid 1980s and the present.
Table 2 shows a regression of the percentage fall in house prices against their previous rise, both including and excluding the Spanish early-1990s outlier, for real house prices and the house price to income ratio. The slope of −0.7 for real house price means that 70 per cent of the rise during a boom (expressed relative to the peak value) is lost during the subsequent bust.
What is notable about the diagram and regressions is how strong the relation- ship between price rises and falls is. Across very different housing markets in very different economies over a period of more than 30 years, there is a common relationship between the magnitude of booms and subsequent busts. Rent-price series show similar mean reversion but because of the small size of the rented sector in many economies, and the presence of rent controls in part of the period, the data are not as reliable as the real price and price-income series.
As always, national averages conceal substantial variations across regions and types of property. During the last British housing crash, for example„ while selling prices nationally fell on average by 10 per cent, they fell in East Anglia by 40 per cent.
As Table 1 suggests, there is a relationship between the magnitude of real price falls and their duration. Table 3 gives the results of a regression of the average annual rate of house price falls on their magnitude, and shows the two to be closely related. If p is the proportionate price fall, so prices fall from 1 to 1 − p over t years, it follows that r = ln(1 − p)/t is the average rate of decline. Table 3 gives the results of a regression of r on p. For every 10 per cent extra decline in real prices, the annual rate of decline rises by 1.5 percentage points.
3 The Irish housing bubble, causes and consequences.
The evidence of nearly 40 cycles in house prices for 17 OECD economies since 1970 shows that real house prices typically give up about 70 per cent of their rise in the subsequent fall, and that these falls occur slowly.
Before looking at what these numbers may imply for Ireland, it is necessary to dispose of the idea that Irish house prices merely reflect strong fundamentals: rising income and increased household formation due to the age structure of the population, declining household size, rising employment, and immigration.
This argument is hard to sustain. If the rise in house prices were due to increased income and more people needing somewhere to live, we would have observed rents rising alongside house prices. Figure 4 shows how house prices have risen far faster than either rents or income. In fact, while rents doubled relative to income between 1995 and 2000, the ratio has remained unchanged since. The failure of rents to rise, along with the number of recently built units that have been bought but are lying empty (Fitz Gerald, 2005), suggests that the Irish housing market has left the dull world of fundamental values far behind it.
A back of envelope calculation of the fundamental price of housing is the following. Abstracting from maintenance costs (which typically run around one month’s rent) suppose that housing generates an annual rent of n. This is a fraction ν of disposable income y which is expected to grow through time at rate g. The present value of this infinite income stream is then
where r is the discount rate. As Figures 1 and 2 and Table 1 show, housing is not a risk-free asset, and this discount rate needs to exceed the risk free rate by an amount reflecting the fundamental risk of the asset. For housing, fundamental rise is large: housing is the largest item by far in most people’s asset portfolio and price changes are strongly correlated with income growth. To be conservative, however, we can assign a value of r of 8 per cent, equal to the long run real return on equities.
The ratio of fundamental price to rent is 1/(r − g). To explain why Irish house prices have doubled relative to rent since 2000 we need to ask if there is any reason to suppose that new information has arrived causing long run estimates of (r − g) to be rationally halved. Ireland’s stagnant exports, diminishing competitiveness, and the increasing structural problems of sectors such as IT and pharmaceuticals, would suggest that estimates of long run income growth for the Irish economy g should have fallen in this period. While it may be the case that increased international demand for quality assets may be driving down equilibrium returns (Caballero, 2006), there is no reason to believe that long run expected returns on risky assets r
have halved in the past 7 years.
As White (2006) has observed, there is considerable variation in price-rent ratios within Dublin, with values in the range 80–100 at the top of the market. These values recall the peaks of the dotcom bubble and can be rationalized, with a dis- count rate r ≈ 0.08, only with real long run growth of income of 6 to 7 per cent, equivalent to a doubling of real income every 10–12 years. This is the rate achieved by Korea during its transition from effectively the stone age to an industrial economy but has not been remotely approached by any rich economy. Alternatively, assuming an equilibrium price-rent ratio in the region of 15, it suggests that large falls in prices, of the order of 85 per cent, might be needed for the top of the market to return to fundamental value.
While other parts of the market appear less over-valued, they are still expensive by international standards. The Global Property Guide website reports that the average Dublin apartment rents for around 4% of its purchase price. Only Madrid among major cities has a lower ratio. By comparison, London apartments return nearly 6%, and Amsterdam and Paris over 8%.
4 International Perspectives on the Irish Housing Bubble.
Were Ireland to experience the same housing dynamics as every other OECD economy, except Spain in the early 1990s, what sort of price changes might be expected? Recall that Table 2 predicts a 7 per cent fall for ever 10 per cent rise (relative to peak values) of real prices from their trough level, with a standard error of 10 per cent.
Since the mid 1990s, real house prices have risen from an index level of 100 to around 350, and increase in terms of peak value of 70%. If seventy per cent of this rise were to be subsequently lost, the predicted fall in real house prices would be 50 per cent with a standard error of 10 per cent. In other words, a 68 per cent confidence interval for price falls would be in the range of 40 to 60 per cent. There would be one chance in eight of a price fall of only 30 to 40 per cent, just as there a predicted one chance in eight of a fall of 60 to 70 per cent.
Similarly, Table 2 predicts, given an approximately 70 per cent rise in the price income ratio, that the price income ratio will fall by around 60 per cent, with a standard error of around 12.5 per cent.
A fall in real prices of 50 per cent from Table 3, implies a predicted annual rate of decline of around 9 per cent, with a standard error of approximately 1.5 per cent. This translates into a decline of around 8 years, of the same order of magnitude as that experienced in the Netherlands in the 1980s or Britain in the 1950s. Assuming an inflation rate of 4 per cent, this implies an annual fall in selling prices of 5 per cent.
These estimates may be unduly optimistic. In all the housing cycles on which the regression was based, housing stock was, for practical purposes, fixed. In Ireland, by contrast, the number of housing units is growing at around 5 per cent per year, which would suggest the potential for larger falls than those experienced in other OECD housing slowdowns.
The prediction that Ireland may experience house price falls in the range of 50 per cent, is a good way from the OECD estimate (Rae and van den Noord, 2006) that Irish houses are overvalued by only around 20 per cent. However, the OECD methodology, and that of similar studies, is problematic. Such studies run a regression of house prices on interest rates, disposable income, employment and other fundamental variables. The regression residuals are then equated with the degree of over- or under-valuation in the market.
To see the difficulty with this approach, suppose that Irish house prices had increased twice as fast as they did, so the regression residuals would double in value. Instead of saying that house prices are over-valued by one hundred and twenty per cent, the residual approach would say that they are overvalued by only forty per cent.
House price falls have three effects. First, households feel less wealthy and consume less. Evidence from the United States points to a final long-run marginal propensity to consume from housing wealth of around 10 per cent: a $100,000 rise in property values, increases household consumption eventually by a total of $10,000 (Carroll, Otsuka and Slacalek, 2006). Secondly, banks face more bad loans, and become more cautious in their lending, leading to further falls in credit- worthiness through the standard financial accelerator. Finally, the value of Tobin’s q for residential investment falls, reducing house building. Most countries devote about 5 per cent of national income to building houses and in a typical housing bust, this falls to around 4 per cent of national income.
In most cases then, housing busts are uncomfortable, but not macroeconomically disastrous events. How about Ireland? There is some evidence that the wealth effect on consumption might not be as strong as in the United States: there has been no fall in personal saving in Ireland during the housing bubble, and households have not consumed home equity through second mortgages. Similarly, the larger banks which dominate lending are well capitalized and the banking system has, until recently at least, avoided the worst excesses of the sub-prime mortgage market, although it is likely that many interest-only and 100 per cent mortgages could go sour, especially given the ease with which delinquent borrowers can relocate to England.
It is the scale of the Irish house building industry that makes a fall in house prices potentially troubling. While most economies derive only 5 per cent of their income directly from residential construction, in Ireland house building accounts for around 15 per cent of national income, with another 3 per cent coming from selling houses.
Effectively, the recent growth of the Irish economy looks similar to the unstable case of an old-fashioned multiplier-accelerator model. The employment growth in the Celtic Tiger period of the 1990s led to increased demand for housing, reflected in rising real house prices and rent to income ratios. This stimulated house building, which generated more employment, leading to more demand for housing, and so on. Effectively, the Irish economy has come to be driven by building houses for all the people whose jobs have come, directly or indirectly, from building houses.
It is hard to envisage how a fall in house building from 18 per cent to 5 per cent of national income might be achieved without considerable macroeconomic dislocation. Building booms, moreover, tend to end suddenly: the example of Arizona in the summer of 2006 shows how a housing market can move in the space of a few months from buyers queuing overnight to buy, to empty tracts of new houses being priced below construction cost and still failing sell.
This paper has taken an international perspective on the Irish housing boom. We have shown that there is a close relationship historically across very different economies and housing markets between the size of increases in real house prices, and subsequent declines. If this relationship were to hold for Ireland, the expected fall in average real house prices is in the range 40 to 60 per cent, over a period of around 8 years. Such a fall would return the ratio of house prices to rents to its level at the start of the decade. Given the unusual reliance of the Irish economy on building houses, the effects of any such fall on national income may be somewhat larger than that experienced at the end of other housing bubbles.
Policy implications are straightforward. Booms and busts are a normal part of property markets. The government did not cause the current boom, and is power- less to do anything about a subsequent bust.
Blanchard (2006) has observed that Euro-area economies appear at risk of ro- tating recessions: increased domestic demand drives up real wages and erodes competitiveness, but the impossibility of devaluing means that prolonged rises in unemployment become the only means to reduce real wages. Notable current ex- amples are Italy and Portugal. There may be some risk that the sharp fall in Irish competitiveness since 2000, which has been disguised and, to some extent, caused by the construction boom, may require a lengthy period of high unemployment to reverse.
Bikchandani, Sushil, David Hirshleifer and Ivo Welch. 1992. “A Theory of Fads, Custom, and Cultural Change as Informational Cascades.” Journal of Political Economy 100:992–1026.
Blanchard, Olivier. 2006. The Difficult Case of Portugal. Working paper Depart- ment of Economics, MIT.
Bulow, Jeremy and Paul Klemperer. 1994. “Rational Frenzies and Crashes.” Jour- nal of Political Economy 102:1–23.
Caballero, Ricardo. 2006. On the Macroeconomics of Asset Shortages. Working Paper 12753 NBER.
Caplin, Andrew and John Leahy. 1994. “Business as Usual, Market Crashes, and Wisdom after the Fact.” American Economic Revieww 84:548–565.
Carroll, Christopher D., Misuzu Otsuka and Jirka Slacalek. 2006. How Large Is the Housing Wealth Effect? A New Approach. Working Paper 12746 NBER.
Fitz Gerald, John. 2005. “The Irish Housing Stock: Growth in the Number of Vacant Dwellings.” Quarterly Economic Commentary pp. 42–63. Spring.
Rae, David and Paul van den Noord. 2006. Ireland’s Housing Boom: What Has Driven It and Have Prices Overshot? Working Paper 492 OECD Economic De- partment.
Shiller, Robert J. 2006. “Long-Term Perspectives on the Current Boom in Home Prices.” The Economists Voice 3(4). Available at bepress.com/ev/vol3/iss4/art4/.
White, Rossa. 2006. Dublin House Prices Heading for 100 times Rent Earned. Technical report Davy Stockbrokers. Availabl e at davydirect.ie/other/pubartic … 060329.pdf.