This site has moved to
The posts below are backup copies from the new site.

December 5, 2007

Robert Shiller: A Psychology Lesson From the Markets

Robert Shiller says the Fed will have trouble reversing downward psychology in housing markets:

A Psychology Lesson From the Markets, by Robert J. Shiller, Economic View, NY Times: It is no surprise that the Federal Reserve’s discount rate cut has not entirely reassured investors. The Fed can stop a run on the banks, but it cannot control the speculative cycle — a cycle built on psychology and misperceptions that has been sweeping much of the world for the last 10 or 12 years.

I have worked with Karl E. Case ... at Wellesley College ... in conducting ... surveys of recent home buyers. In Los Angeles and San Francisco in 2005, when actual home prices were rising more than 20 percent a year, we found that respondents anticipated big increases far into the future..., the median expected annual climb for the succeeding 10 years was 9 percent.

This expectation would mean that a house valued at an already high level of $650,000 in 2005 would be worth more than $1.5 million in 2015. For most people in 2005, it would also mean that they should buy a house soon, or forever be excluded from owning one — and that it would be better to stretch and buy the most expensive house they could afford...

Now, of course, prices have been falling, and our survey over the last few months shows that in Los Angeles and San Francisco, the median 10-year expected price increase among recent home buyers has come down to 5 percent a year — a number that is likely to decline further if prices continue to drop. As price expectations fall, homeowners lose the incentive to pay off a mortgage on a home they are realizing is beyond their means. They decide to default. We thus have the beginnings of a mortgage crisis.

The problem is fundamental, tied to the imbalance caused by irrationally high home prices and declining optimism that the prices will go higher. Cutting interest rates will not change this basic situation. The problem is fundamental because the speculative cycle afflicts much of the world. ...

Classical economics cannot explain this cycle, because underlying these booms is popular reaction to the price increases themselves. Rising prices encourage investors to expect more price increases, and their optimism feeds back into even more increases, again and again in a vicious circle. As the boom continues, there is less fear of borrowing heavily, or of lending heavily. In this situation, lower lending standards seem perfectly appropriate — and even a fair way to permit everyone to prosper.

Booms cannot go on forever. Downward price feedback sets in. That is when balance sheets become impaired and widening credit problems start to show up.

The puzzle is why this speculative cycle has occurred recently in so much of the world. What do all these countries share that drives them to speculative booms? ...

As we all try to adjust to a rapidly growing and increasingly capitalist world, we have been trying to discover who we are and how we fit into it. This has meant an enormous change in values.

Many people feel that they have discovered their true inner genius as investors and have relished the new self-expression and excitement. Investors across the world have been thinking that they are winners — not recognizing that much of their success is only a result of a boom. Declines in asset prices endanger this very self-esteem.

That is why it is so hard to turn around investor attitudes once a downward psychology sets in. The Fed and other central banks do not have lithium or Prozac in their bag of remedies, and so cannot control it.

I don't have much to say in response to this, maybe you can come up with something, but let me point you to a much more detailed discussion of this topic by Shiller:

Historic Turning Points in Real Estate, by Robert Shiller, Cowles Foundation Discussion Paper No. 1610, June 2007: Abstract This paper looks for markers of ends of real estate booms or busts. The changes in market psychology and related indicators that occurred at real estate market turning points in the United States since the 1980s are compared with changes at turning points in the more distant past. In all these episodes changes in an atmosphere of optimism about the future course of home prices, changes in public interpretation of the boom, as well as evidence of supply response to the high prices of a boom, are noted.

Introduction By some accounts, the greatest challenge for economic forecasters is to predict turning points. It is easy to extrapolate time series. It is less easy to tell when the series will abruptly change trend and enter a different pattern or regime.

Figure 1 shows a chart of US stock price and home price indices since 1987. It shows the Standard & Poor 500 Stock Price Index... It also shows the Standard & Poor/Case-Shiller Composite Home Price Index, a ten-city average which is a measure of the aggregate market for single family homes, and is based on indices that Karl Case and I created in 1988. ... Since the home price index is a threemonth moving average, the S&P 500 is also plotted as a moving three-month moving average so that the two series are comparable.


The eye naturally picks out what appear to be major historic turning points. Looking at the chart, one sees that the stock market has shown abrupt changes in regime in early 2000, when it left a strong bull market, and went into rapid declines, and in late 2002 or early 2003, when it resumed a new strong upward climb. The real estate market changed its direction markedly around 1990, from a booming market to a market in the doldrums for the better part of a decade, and then the market started accelerating upwards at increasing rates. The national home price boom since the late 1990s appears unprecedented in US history, although the “baby boom” in housing of the late 1940s and early 1950s comes close, and there have been some very large local booms. The rate of US housing appreciation slowed after 2005, and, to some eyes at least, it would appear just sometime after mid 2006, we are entering a new regime of downward price changes.

These, then are the natural questions for today: How shall we think about these major turning points of the past? Can study of these turning points offer to us any way to predict when the upward trend in the stock market? Is there any way to decide whether we really are entering a new regime of real estate market price declines? Some people who think in terms of time-series analysis may disagree that such questions should even be asked. Prices observed in the stock market are widely described as random walks, if only approximately. If a time series is a true random walk, of a kind that may be generated by a random number generator, then it will be seen to have occasional major “turning points” that one could say have no other explanation than the chance arrival of a string of negative shocks after that point. Probability theorists can calculate the probability that the random walk will surpass the peak again, or calculate the improper spectral density of the random walk, but they would not seek to “explain” the turning point.

Of course, even if an economic time series is found to have the stochastic properties of a random walk, we could still “explain” the turning point by interpreting the sequence of shocks to the random walk that allow us afterwards to choose a point as the turning point. We might be able to tell a story about the causes of the sequence of negative shocks that came afterwards, which of course were really not generated by a random number generator but instead have interpretations in terms of various historical events. But, so long as the series is truly a random walk, the explanation would have to be entirely after the fact, and would offer no insights into the future forecasting of turning points.

Stock prices are not known to be exactly random walks. It has been demonstrated, for example, that stock prices have shown some momentum through time. Certain models very different from a random walk, involving such things as sudden regime changes, are not easy to reject statistically. There is a substantial econometric literature that documents deviations from random walk properties, and there is an econometric literature on the identification of regime change in time series. But stock prices are fairly close to random walks.

While stock prices somewhat resemble random walks, real estate prices certainly do not. There is, in fact, a very obvious smoothness to home prices historically, as can be seen from Figure 1. In fact, if one fits a quartic polynomial to the home price series shown in the figure one gets an R squared, as a measure of closeness of approximation, of 99.6%. Of course one should not use the fitted polynomial as a forecasting device, but the goodness of fit does illustrate the smoothness of the series, which no doubt has something to do with the difficulty that professionals and speculators have in reacting to new information about the housing market. The housing market is populated mainly ordinary folk who do not react with the speed of professionals. 


A polynomial has no unambiguous turning points. The point at which the slope changes from positive to negative or negative to positive may stand out to the eye on a plot of the polynomial, but the same point would not stand out on a plot of the slope of the derivative of the polynomial. Indeed, Figure 2, which shows the annualized rate of change of the same monthly series that appears in Figure 1 along with the National Association of Home Builders Index of Traffic of Prospective Buyers, gives a somewhat different impression where turning points might lie. There is of course seasonality, hard to detect on Figure 1, but which stands out as a powerful annual oscillation in Figure 2. Beyond that, it is clear that the housing market goes through long periods of either steadily increasing or steadily decreasing home price inflation, and that these periods have ended rather abruptly.

For example, it would appear that there was a major turning point in 2004 when home price increases peaked, and that corresponds roughly to a time when the traffic of home buyers also peaked, something that would probably stand out to home builders and real estate agents as the true turning point. The plot of increases in home prices in Figure 2 looks rather more (abstracting from seasonality) like a broken straight line composed of about three line segments. The rate of growth of home prices made abrupt changes in trend in 1991 and 2004. Because of the suddenness of the change in growth rates, these dates are only a couple years away from the breakpoints indicated by looking at the Levels chart, 1989 and 2006, in Figure 1. ...

I want here to pursue the “why” of the apparent turning points, linking the apparent regime changes to economic events and to principles of behavioral economics. I will not be able to give a complete answer, as I will not be systematically pursuing all economic factors that might logically have an impact on home prices. On the other hand, in this paper I will concentrate on some apparently important psychological factors that are likely to be omitted completely in a rigorous econometric analysis of the real estate market.

The method here will be largely narrative, recounting the stories that people told about the market of the times. Economists are usually very careful to avoid entering such evidence. And yet, research by psychologists has found that narrative-based thinking is extremely important in human decision making. People’s thinking is often more influenced by human-interest stories than they are by quantitative evidence, see for example Schank and Abelson (1977,1995). Pursuing such an analysis may actually help us to forecast turning points, providing information that is hard to pursue with rigorous econometric analysis, or that may at least augment such econometric analysis by suggesting alternative models or suggesting priors for models. ... [continue reading here]

No comments: