Saturday, March 31, 2007


How vulnerable is the United States economy to fluctuations in the housing market? Both historical and more recent statistics indicate that the well being of the United State economy, measured by GDP and median income, is married to the housing market. A further examination of United States GDP and median home prices shows an increased degree of correlation in periods immediately after excessive growth in median home prices.


Correlation is simply the strength of the relationship between two variables. For example, when the sales of snow skis go up, the sales of ski boots are likely to go up by a similar degree. The two could be said to be highly correlated with each other. Furthermore, if one knew the sales trends for snow skis, one could make reasonable assumptions about the sales of ski boots. The correlation coefficient in simply a mathematical way of representing the correlation, or the strength of the relationship, between two variables. As the correlation coefficient approaches 1.0, the correlation of the two variable increases. Two variables with a correlation coefficient of 1.0 would move perfectly synchronously with each other.

Historical Correlation

Historically there has been significant correlation between United States median home prices and GDP. Correlation coefficient between the GDP Quantity Index and median sales prices of existing homes is been .9631.

The correlation between GDP and median home prices becomes more apparent if we examine the rate of growth in each of the indexes. The correlation between growth in GDP and home prices increase immediately after periods of extreme growth in median housing prices.

Fundamentally this correlation makes sense. As home prices decline:

    • Individuals may be less likely to buy a new home for fear that home prices may continue to decline, decreasing demand for real estate brokers, construction workers, mortgage lenders, other housing oriented jobs and other employment related directly and indirectly to housing.
    • Reduced employment translates into less consumption spreading to other business segments of the economy causing further unemployment; and the cycle then continues to recession.

These fundamental phenomena also help to explain why (1) spending on home improvement and maintenance and (2) median household income are highly correlated with both GDP and median home prices (a comparison that we will later examine)

There is also a high correlation between Employment and Housing Prices as reflected on the following chart. The correlation coefficient between the employment and median sales prices of existing homes has been .9479

Increased Correlation During Times of Extreme Growth in Housing

The correlation between GDP and median home prices is even more noticeable immediately preceding period of excessive growth in real estate prices. In both 1980 and 1987 we saw national median home price growth approach or exceed 10%. In both scenarios, that growth was succeeded by a rapid slowing and eventual decline.

Delta for GDP vs. Delta for Median Home Prices from 1981 to 1983

This graph shows that the rate of increase in housing prices peaked in the 2nd quarter of 1981 and the rate of growth in GDP peaked the next (3rd quarter) of 1981. The chart also shows that housing prices starting increasing again in the 3rd quarter of 1982 and that GDP also started increasing at the same time.

Interestingly, look at the amazing similarities in the following chart for employment vs. median homes prices for the same period.

Delta for Employment vs. Delta for Median Home Prices from 1981 to 1983

Similarly; look at the period 1987 to 1991.

Delta for GDP vs. Delta for Median Home Prices from 1987 to 1991

In the 3rd quarter of 1987 growth in median home prices approached 10%. Growth in home prices began to decline drastically reaching -2% in the 4th quarter of 1990. This drop in home price growth was accompanied by a precipitous and highly correlated drop in the growth rate of GDP, which dropped below 0% at about the same time that home prices reached -2% in the 4th quarter of 1990, and continued to drop, reaching -1% in the 1st quarter of 1991.

Again look at the chart for employment vs. Median Home Prices.

Delta for Employment vs. Delta for Median Home Prices from 1987 to 1991

The phenomenon of increased correlation between median home prices and GDP indicates the national economy’s increased reliance on the housing market during times of extreme housing appreciation. As the real estate market booms, the national economy morphs to accommodate growth in the real estate sector. The real estate sector begins to attract a larger portion of jobs as more construction workers are needed to build homes, more real estate brokers are needed to sell homes and more mortgage brokers are needed to help finance homes. The real estate sector begins to drive a larger portion of spending as people spend more money on new homes, spend money acquired from home equity lines of credit secured by the appreciated value of their home, and spend more money on home improvement in preparation of a sale or to make over a newly purchased home. As the real estate boom subsides, the economy loses a source of employment and spending that it has exhibited increased reliance on. As such, declines in the real estate market subsequent to a large real estate boom have been highly correlated with a sharp decline in GDP.

Recent Years

Recent times have brought growth levels in the real estate market that bear a striking resemblance to that of the early and late eighties. Growth in median home prices in the third quarter of 2005 was 14.99%, the highest in recorded history.

This growth in the real estate sector has resulted in a correlated stable growth in GDP. How much of the growth in GDP is attributable to growth in the real estate market? In recent years consumer spending and residential construction have accounted for 90% of GDP growth. And employment in housing and related industries accounted for 43% of the increase in private sector payrolls since November 2001. Combined with the historical correlations of declines in GDP following real estate booms one must consider that if the real estate sector were to enter a down cycle, GDP would likely be dragged down with it. According to the 2006 UCLA Economic Forecast, if real spending on housing returns to its pre-2000 peak, the resulting impact on GDP would be a reduction of $150 billion (1% of GDP).

Individual Cities

One common belief espoused by the real-estate-bubble-nonbelievers, is that if a real estate bubble does exist, it is isolated to a few major metropolitan areas and thus a burst in the bubble is unlikely to cause wide spread problems for the national economy. There may be some truth to the contention that the real estate bubble is isolated. Median home prices in some isolated areas like Miami and Los Angeles have reached annual growth levels approaching 30%! However, as we will see, when discussing the effect of a downturn in a real estate market on the national economy, a focus on isolated bubbles is all that is necessary.

This chart illustrates the growth of median home prices in 14 large metropolitan areas relative to growth in median home prices nationally.

(click for a larger version) Median Home Price Estimates

Whether there has been a bubble in the national market or simply in more local markets lies in the definition of a “bubble”. Relative to historical standards there has been exceptional growth in national median home prices. However, the growth in certain isolated markets is even more extraordinary, approaching four times the national level in certain areas like Phoenix. Some might say that there is no national real estate bubble, only localized real estate bubbles. Others might look at the statistics and say that there is a national real estate bubble and localized real estate super bubbles.

Compare recent growth of home prices in Phoenix, AZ to growth in GDP and the results are astounding:

GDP vs. Median Home Prices in Phoenix

While the growth in Arizonian median home prices may seem like an extreme example, they are only a mild exaggeration of the average among these large metropolitan areas.

GDP vs. Average Median Home Prices Among 14 Large Metro Areas

These large metropolitan areas have experienced exorbitant growth in recent years. The non-real-estate-bubble believers may have a point to the extent that they espouse the belief that there are isolated pockets of extreme growth. However, the extreme growth in these markets does not negate the still unusually high growth nationwide. Furthermore, the position of the real estate bubble naysayers maintaining that the real bubbles are isolated to a handful of metropolitan areas is often premised on the belief that the real estate economy in those regions could take a turn for the worse without significantly affecting the national economy. As we will see, when discussing the national economy, to say that there is no national real estate bubble, only localized real estate bubbles, is akin to saying the airplane is not on fire, only the wings are on fire.

Correlation of Median Home Prices in Top Metropolitan Areas with GDP

How vulnerable is the national economy to the real estate economy of a handful of metropolitan area? If we examine median home prices in the fourteen large metropolitan areas listed above we find that the well being of the national economy bears an even higher correlation to the cumulative changes in real estate prices in these fourteen cities than it does to national home prices.

Since 1968, the correlation coefficient of GDP to national median home prices is .9631 vs. a correlation coefficient of .9766 of GDP to the average median home price of these fourteen cities. This is a strong indication that changes in the median home prices in these fourteen cities will actually have a more powerful effect on our national economy than a change in home prices nationwide.

There are a couple of fundamental factors that might explain this phenomenon of high correlation between home prices in these fourteen cities and the national economy.

First, the large metropolitan areas host the nation’s most expensive homes. These homes are apt to lose the most value in time of real estate slow down. The inexpensive homes in more rural areas are likely to have more stable prices. Essentially, areas with higher median home prices are likely to experience greater volatility. For example, over the past 35 years, Dallas has had an average median home price of $93,000. The standard deviation of home prices in Dallas is $35,000. Los Angeles’ average median home price for the same period was $152,000, about three times that of Dallas. The standard deviation of home prices in Los Angeles is $95,000. Stated differently, the standard deviation to average median home price ratio is .37 for Dallas and .63 for Los Angeles. In proportion to its home prices, Los Angeles is almost twice as volatile as Dallas.

Second, because these fourteen cities are so large and because their median home prices are so high, fluctuations in the home prices in these cities result in an increased net economic effect. For example, a $150,000 decrease in median home prices in Los Angeles will have 10 times the effect of the net value of our national housing stock than a similar decrease in a city that is one-tenth the size. And because median home values in Los Angeles tend to be higher, a 10% decrease in median home prices in Los Angeles will represent a greater net decrease in median home values than a 10% decrease is median home values for a less expensive market. Thus, because these large cities have (1) more expensive homes and (2) more homes, decline in home prices of any given percentage in these cities will tend to have a greater impact on our national economy than similar changes in smaller markets.

Third, fluctuations in values of expensive homes can have a more direct effect on consumer spending than fluctuations in prices of less expensive homes. As expensive homes drop in value less money is spent on construction, less real estate commissions are paid and less mortgage dollars are issued. Further, as wealthy individuals lose equity in their expensive homes they are highly likely to forego luxury good purchases. When a wealthy individual loses $1,000,000 of equity in his home he can forego the purchase of his weekly dinner at Spago’s. On the other hand, an individual of more moderate means is unlikely to purchase less canned tuna because his equity in his home declines. This may be another reason that home prices in more expensive areas tend to bear an increased correlation to GDP.

The high median home prices in these fourteen cities thus become a triple edge sword wielding powerful correlation with our national economy. As discussed above, home prices in these fourteen cities is a superior indicator of the direction or our national economy that national median home prices. The calculation of national median home prices is mathematically inefficient at representing the effect of home prices on the national economy and the true impact of home prices in these fourteen large cities on the national economy is misrepresented in national median home prices calculations.

The inaccuracy of national median home price calculation at estimating the effect of the housing market on the national economy stems from the fact that national median home price calculations tend to ignore much of the volatility in these large regional markets.

When correlating home prices to GDP we find that median home price statistics tend to underestimate the effect of home prices on the economy. Median home price statistics seek to estimate the price of a “typical” property and thus choose the price of the home that is in the middle of the sample so that 50% of the samples have a price above median home price and 50% of the samples have a price below median home price. Thus, if one hundred expensive homes worth $50,000,000 each drop to $2,000,000, the median home price remains unchanged even though the value of the nation’s total housing stock has dropped $4,800,000,000. (This assumes the median home price is below $2,000,000) Average home prices on the other hand simply take the cumulative value of all homes and divide that value by the number of homes. Thus, the calculation of average home price would take into account a drop in the price of expensive homes even if inexpensive home prices remain stable. This may seem like an extreme and unlikely example, but it highlights the differences in the mathematical calculations between median and average. Median home prices will tend to underestimate the effect of fluctuations in home prices above the median to the extent that there is not an equal fluctuation below the median.

Thus; since for instance, the median priced home in Los Angeles is significantly higher than the national median priced home; Los Angeles could experience a significant drop in prices, however, if Los Angeles has few homes that sell below the national median priced home; little, if any of the drop in prices in Los Angeles will be reflected in any change national median home prices.

Now, knowing that (1) home prices in expensive areas are more volatile than those in inexpensive areas and (2) median home price statistics are likely to underestimate that volatility, we deduce that average home prices should be more volatile than median home prices. The following chart illustrates this phenomenon:

National Association of Realtors: Real Estate Outlook; Adjusted

It is important to recognize the difference between median home prices and average home prices. While median home prices may be a good indicator of the affordability of homes for the average American, they may not accurately represent the effect of home prices on the economy. This is substantiated by a high correlation coefficient of .989 between average home prices and GDP for the period from 1968 to 2005.

So why do we still discuss median home prices? Median home price statistics remain a more popular method of analyzing price fluctuations in the housing market because they are said to better represent the affordability of homes for the average American. They are more widely published and have been better documented. They will also be more familiar to the reader. As such, median home prices remain the topic of this paper. But one should be aware of the difference between median home prices and average home price and the distortive effect of median home price statistics. The ultimate point of this entire discussion of median vs. average home prices is that nationwide median home prices do not accurately represent the effect of fluctuations in prices in higher priced areas. However, the changes of home prices in those high priced areas bear a high correlation to changes in GDP.

The following chart illustrates the changes in GDP and the average of the median home prices for the fourteen cities listed above.

Note the precipitous drop of the changes in home prices starting in the 3rd quarter of 1981 followed the next quarter by a drop in GDP; all of which later culminated in a recession beginning in the first quarter of 1982.

Note the continuous drop in the change in home prices starting in the 2nd quarter of 1988 which slowly starts being reflected in slowing GDP starting in the 2nd quarter of 1989 and continuing to recession in the 1st quarter of 1991.

Note the decrease in change in prices from the last quarter of 2000 to the last quarter of 2001, matched by a decrease in GDP. Further note, thereafter, the upward trend in change in prices through the 2nd quarter of 2004 matched by an upward trend of GDP. Finally, note the continued increase in the change in prices during 2005 that was only matched by a general leveling of increase in GDP.

In the past when growth in these cities has approached or exceeded 16%, there has been a subsequent precipitous decline in home price growth, followed by a period of negative growth. These periods of negative growth in home prices were accompanied by a period of negative growth in GDP.

Home prices in some large metropolitan areas may indeed be significantly more inflated than the national median. However, home prices in those areas also bear a higher correlation with GDP. Whatever happens to home prices in these fourteen cities following an extreme increase in prices would, based on the two previous recessions, indicate the direction of the national economy.

As real estate prices in those areas decline, so does GDP.


Based upon the foregoing it makes sense to look at home price growth rates in the past. Perhaps the rates we have been experiencing of recent times are high; but not high by historical standards.

First, let’s look at national prices. In both 1980 and 1987; the growth in home prices preceding the last two real estate lead recessions, growth rates were less than the recent peak of ___%. This certainly, in historical terms should draw serious concern.



Next, let’s look to our 14 cities and see how they are behaving. We find that those cities also are growing at rates in excess of anything in recent history. It is also quite clear from both the chart national and 14 city charts that periods of significantly reducing growth rates followed the two pervious occasions when growth rates climbed to such high levels.

Average of Fourteen Cities

Further, there are some cities that are particularly hot where we might expect some even more drastic type response. Look at the following graphs of Los Angeles, Washington DC, Jacksonville, Miami and Phoenix. These cities are not typical of extreme price growth happening around the US; however, these are highly populated areas with significant economies that cannot be ignored with respect to the impact they can have on the national economy. The growth rates experienced in these cities is astonishing and cannot be sustained.

Having looked at all these graphs; we should keep one thing in mind; a person should be careful not to misuse statistics the way a drunk misuses a lamp post – for support rather than illumination. We derive from the foregoing charts that house price growth has been excessive and that in the past when this has occurred it has been followed by a period of slower growth which has lead to recession. While we will review a lot of other data and charts through out this book; ultimately; we must look to other support to draw any final conclusions.

Correlation Coefficient indicates the strength and direction of a linear relationship between two random variables. In general statistical usage, correlation or co-relation refers to the departure of two variables from independence. (

Northern Trust Company Daily Economic Comment, May 23, 2005.

2006 UCLA Economic Forecast

Tuesday, March 20, 2007

Psychology of a Housing Bubble

Question: How can a tulip bulb become worth $76,000?

Answer: When a buyer is willing to pay that much.

The price of an asset in a competitive market, economists tell us, occurs at the equilibrium level where the supply curve intersects the demand curve on a graph where price level lies on the y axis and quantity demanded lies on the x axis. This pricing theory, so elegant and simple, is the foundation of every course in economics, and is accepted as a universal economic truth that concisely explains consumer behavior in all corners of the free market. It is said that this model can be used to explain and predict changes in the price and quantity of goods sold, but unfortunately the usefulness of this model is limited by our ability to accurately ascertain the necessary inputs. The model has failed miserably in the past because we have no accurate method of drawing the supply and demand lines on the graph with any degree of scientific accuracy. Mathematical modeling does not determine where the demand (and supply) lines are to be drawn. Consumer (and supplier) behavior determine location and slope of any line, and unfortunately for economists, human behavior is not influenced by graphs, equilibrium models, and demand theories.


At the peak of the market In Holland in the1630’s, tulib bulbs sold for as much as $76,000 (present day dollars). Six weeks later the same bulbs traded for less than a dollar. Economic modeling cannot predict or explain such pricing phenomenon because economic models make a critical assumption about human behavior that is rarely an observable truth. Economists assume rational consumer behavior in order to make future predictions. History has proven that nothing could be less rational than making such an assumption.

To explain buyer behavior in any market where prices are changing requires careful consideration of the psychology of the market. Irrational and manic behavior influences every market, and the result is that prices change for reasons that cannot be explained without considering the ever-changing beliefs of the market’s participants. The story of Tulips in Holland is merely one example of how market psychology can change without any fundamental changes in the underlying market, and cause ruthless damage to those too deeply entrenched in the mania.

The story began in 1593, when Conrad Guestner imported the first tulip bulb into Holland. Initially the rare bulbs were purchased by the wealthy as a sign of status and wealth, but soon thereafter speculators entered the market seeking to turn a profit, and as demand increased along side with trading activity, speculators earned very healthy profits. Tulips began to be traded on market exchanges, and by 1634 trading was conducted mainly by people in the middle class looking to make their fortunes in a market where prices steadily rose. Soon the entire Dutch nation, hypnotized by the success of the market, were selling their farms, livestock, and life savings to speculate on a single tulip bulb, and prices rose astonishingly quickly, with few people pausing to consider the rationality of the market’s behavior. Trading activity quickly spread across Europe to other exchanges, where tulip derivatives (in this case option contracts and futures contracts) allowed even greater speculation by broader cross sections of the population.

Then came a seemingly insignificant event. The Dutch government began issuing warnings and started to develop regulation to help control what they believed was dangerous price appreciation. This prompted some speculators to cash out, believing that prices tomorrow could be lower than they are today should such regulation be imposed. Furthermore, another curious event happened. People grew new tulip bulbs, and tulips weren’t quite as rare any more. Soon thereafter, tulips began a slight downtrend, and all the speculators began to panic at the same time – unfortunately panic spreads like a virus, just much more quickly. With remarkable synchronism, the speculators’ had the identical reaction to lower future price expectations – sell the tulip bulbs today. The resulting price crash, possibly predictable by a psychologist or behavioral expert, could not have been predicted by an economic algorithm. The Dutch government tried to engineer a bailout, but the problem was too massive to prevent large numbers of bankruptcies, defaults, and lost fortunes. Bulbs were worthless, and the Dutch economy was crippled for decades.

Feedback Loops

What caused the rapid crash of tulip prices has become known as a “feedback loop”, where investors react to price declines resulting from a minor sell-off by selling themselves, leading to greater price declines, and further selling by other investors. Robert Schiller, in “Irrational Exuberance” relied on this theory of investor psychology to accurately predict the stock market crash in 2000[1]. Schiller analyzed the 1987 stock market crash and, where he concluded that feedback loops precipitated that crash, a conclusion supported by his own research and by the Brady Commission study conducted in the wake of the crash.

It is important to recognize that “feedback loops” is a psychological theory unrelated to pricing fundamentals, such as earnings or interest rates. Psychological theories about asset class price declines are considered by about 2/3 of all institutional and individual investors as being more important in predicting price declines than fundamental pricing methodologies[2]. This belief of most investors has far reaching consequences, as any student of Game Theory will soon recognize. At the heart of Game Theory is the concept that investors make their own investment decisions based on their expectations of what decisions other investors will make. If 2/3 of investors believe that psychological phenomena play the most important role in the pricing decisions of other investors, then the importance of any clues relating to the psychology of the market become immediately apparent. The research suggests that the most successful investors may be those who are the best psycho-analysts for the market, more than they are experts in predicting the future fundamental performance of a market’s underlying assets.

Predicting Market Psychology Today

The tulip mania, the 1987 stock market crash, and the 2000 dotcom bust are merely illustrative of how consumer psychology can lead to irrational behavior and unsustainable and unpredictable changes in price levels. When consumers believe that an asset class will appreciate, be it tulips, tech stocks, commodities or real estate, irrational behavior can lead to irrational prices. The ability to identify irrational behavior early is essential to prosperity in any investment market, as irrational prices always correct themselves at some point.

The key to identifying irrational behavior is to identify euphoric public sentiment about an asset class, accompanied by manic speculative investment activity, and rapidly rising prices. When these three market characteristics align, this is normally a sign that what will follow is decreased expectations regarding the future value of the asset class as some speculators begin to sell of and capture profits, while others introduce new supply. These events trigger the belief that asset prices will be lower tomorrow than they are today, and such thinking explodes across the investment community, and prices correct, or “mean revert”.

Today, we see some evidence of manic public sentiment about real estate. Some notable proof of this sentiment can be found in:

  • The proliferation of seminars prosthelitizing the benefits of investing in real estate. One of the most notable of these seminars is hosted by the icon of many wannabe millionaires, Donald Trump.[3]

  • A reality TV show titled “Flip This House” where an investor buys a house, remodels it and tries to sell it for a profit.[4] Their website even include a section titled “Flip Tips” where they give the viewer tips so they can go play the game on their own.[5]
  • currently has 1694 books listed in its real estate investing section[6] This notable publication is the 6th most popular:

  •, a site dedicated to helping people find real estate, real estate agents and mortgages is one of the top 60 sites in the United States, beating out,,,,,, and of course,[7]

The problem of course is that market psychology is not as easy to predict as I have just described. The best we can do is compare the psychology of the current market to other periods in history where time has proven that consumer behavior was irrational. It may be helpful to read the above history of the Tulip craze one more time, this time comparing the story to today’s real estate investment environment. Today’s real estate investors may be well advised to do the same for other periods of investment irrationality and compare the characteristics of those markets to those of the real estate market today. If the analogies strike the investor as somewhat scary, then this could be a valuable insight into the market that might help him capture any gains he has achieved before losing them. If the investor still believes concern is unwarranted for the time being, then this could confirm his bullish view and allow him to move forward with greater confidence and certainty about his position.

Triggers of Shifting Market Sentiment

If one can accurately predict a forthcoming shift in investor sentiment related to an asset class, then profit potential is unlimited. Unfortunately predicting such a shift with accuracy is impossible. However, investors do have some clues that can be used to shift the odds his favor. Human behaviour experts submit that the brain merely responds and reacts to inputs received from the body’s sensory organs. On a micro level, we cannot expect to understand any individual investor’s mechanisms for processing inputs and formulating appropriate responses, but on a macro level we can look to the inputs that are available to the market, and make assumptions about what reactions we can expect based on those inputs. To go one step further, we can try to evaluate which are the inputs that the most number of investors are exposed to, which may help us predict the responses of large groups of investors, and hence broad market moves.

Fortunately the many of the inputs available to investors are not proprietary. Most investors rely heavily on news reports, investing journals, publicly released market data, government reports, and expert analysis, all of which is available to anyone willing to look. This mode of analysis is by no means new – one of the most popular methods by which historical market moves have been analyzed is by reference to the media accounts and published reports available to investors at the time the shift occurred. Over 20% of Robert Schiller’s “Irrational Exuberance” was devoted to discussing the impact of the news media, and investor’s responses to the publications of the time. In this regard we consider trends in the media over time, alongside investor reactions to media publications in the form of price movements in the real estate market. We assume for purposes of this analysis that investors react to inputs from the media and making buying and selling decisions accordingly as was discussed previously in this chapter. It is up to the reader to evaluate the validity of this proposition, but we believe there is ample evidence suggesting that inputs received from the news media plays as significant role in the psychology of a market’s participants for the reasons discussed above.

[1] Irrational Exuberance, p. 90

[2] id.