Ethical Homes

A lot of people have been asking me recently about this article in The Economist, which goes to great lengths to imply (but never say outright) that there is a global housing bubble, and that a worldwide housing crash is imminent.

The short version of my answer: I can’t think of a more intellectually dishonest article that I’ve read in ages.

First, let me make clear that I’m not saying that there is a housing bubble, and I’m not saying that there isn’t a housing bubble—all I’m saying is that the Economist presents what appears to be an intentionally very shallow discussion of the housing market, which I can only assume was done in order to be able to allow them to imply a more dire conclusion that would, presumably, sell more issues of their magazine.

I’ll get to my views on a possible housing bubble at the end of this article, but before I do so, I think it’s important that people understand why, in their article, The Economist reveals itself to not be a credible source of information. It’s what I call the Michael Moore effect. I detest Moore, not because of his politics (much of which I often agree with), but because of his blatant disregard for the truth, and his need to present as un-balanced picture as possible when making a case. The best example might be the scene in Bowling For Columbine where he contrasts numbers of gun-related deaths in the US vs. other countries—and does so in absolute numbers, rather than per-capita ones, claiming that the annual gun-related death rate in the US is 11,127, vs. 381 in Germany. When corrected to reflect the fact that Germany has far fewer people, the relevant numbers are actually that the US has an annual gun-related death rate of around 38 people per million, while the rate in Germany is around 4 per million; that disparity is still more than significant enough to have made Moore’s point, but he would clearly rather make the most extreme case than actually treat his viewers (and the truth) with any respect. In the process he loses all credibility—and that’s exactly what, IMHO, the Economist does in its article as well, by both intentionally manipulating the presentation of numbers to suit their thesis, and by failing to provide any sort of reasonable model that explains their fear-mongering conclusions.

Manipulating The Numbers

Like Moore, the authors of the article in The Economist pick and choose the numbers that suit their thesis. When discussing the end of the housing boom in Australia, for example, they point out that while the official statistics show that that market has had a “soft landing”, with annual appreciation going from 20% to less than 1%, an alternate set of statistics from the Commonwealth Bank of Australia (based on when contracts are signed rather than when those contracts go to settlement) show current values dropping. Analyzing prices at the time of contract ratification rather than settlement does actually make sense, as that is when the negotiations happen and the market decides the then-current value of the property. But the very fact that an alternate set of statistics exists that proves the exact opposite point from the one made by the official government statistics should emphasize the fact that the conclusions reached in an analysis are entirely dependent on which set of statistics is chosen. Any analyst worth their salt, then, is required when choosing a particular statistic to explain why that statistic is the most relevant one; in particular when choosing a statistic that is not the generally accepted one, moreover, an analyst has to explain what other deviations from the accepted methodology are possible, and why only certain ones were chosen.

I’m fairly certain that both the Aussie government stats and the Commonwealth Bank stats use the prices recorded in the public records, for example; those prices, however, don’t necessarily reflect the true cost to a buyer of purchasing a particular property. In the Metro DC market right now, some sellers are (foolishly) forgoing agents, but most buyers are still using agents; if the sellers don’t then allow the buyers to fold the fees charged by the buyer’s agent into the sales price of the home (so that the buyer can finance those fees as part of their mortgage), then those fees don’t show up in the public record sales prices, which as a result under-represent the effective cost of those properties. This might seem to be irrelevant to the decision of whether to use the settlement-date stats or the contract-date stats, but consider this—what if sellers who don’t use agents take longer to get to settlement than sellers who do use agents (because the agents proactively prevent more potential issues that could delay settlement), or take less time to go to settlement (because the agents insist on a slower, more deliberate process than a layperson in order to ensure that all of the appropriate steps are taken)? If the Australian market is similar to the DC market in the prevalence of FSBO sellers, then either situation would change the effect of the switch from settlement date to contract date as the basis for statistics, which could in turn actually make the contract-date stats less reliable than the settlement-date stats, despite the fact that contract dates are, in fact, more closely tied to true market values. Without discussing these other potential factors affecting the price, and how the Commonwealth Bank study may or may not have corrected for them, the second set of statistics has NO reason to be viewed as more accurate than the first one—but The Economist needs to make their point, so the second one is set out as implicitly more accurate.

Again looking at their Australian statistics, they contrast a current 0.4% “12-month rate of increase” in prices according to the government statistics with a 7% drop in prices according to the Commonwealth Bank statistics—but that 7% drop is “since 2003″, meaning the annual drop is only 3.5%; as with Moore and the death rate stats, the real statistics still make their point, but they would rather present those statistics incorrectly in order to make that point seem more important. And, again, it’s worth emphasizing that without more information about how those alternate statistics were calculated, they simply can’t be compared with the standard statistics; it’s possible that the Australian market is experiencing a drop (although a 3.5% annual downward correction is by no means a bursting bubble), but it’s just as possible that that market did have a soft landing, and by presenting such a biased representation of the statistics, The Economist gives us no reason to believe their interpretation over any other.

They manipulate the reader in many other ways, but I don’t have the time to dissect each one; instead, I’ll just show what I think is the most egregious example: the manipulative use of graphs. The coup-de-grace in their “argument” for the existence of a bubble is a comparison of the US, British, and Australian housing markets with that of Japan, which has been in a serious downturn for the last 14 years:

This is a pretty miserable graph, for a variety of reasons: there are no units given for the Y axis, for example, without which the graph conveys absolutely no information of value; it’s also hard to take seriously a graph whose legend claims that for the Japan line, the value in 1985 was 100, even though the graph clearly has the Japan line starting at 100 in 1980, and climbing to around 140 by 1985. Leaving those technical flaws of presentation aside, though, let’s look at the implied point of the graph—that western housing markets are now at the top of a peak that parallels very closely the peak of the Japanese market. Let me present in counterpoint a similar graph that I just whipped up, comparing the Japanese market with the US market over the 30 years that the US OFHEO has been tracking housing prices:

The US data in my graph is from the OFHEO 2005 first quarter housing price index data set. For the Japanese market data, I am using a rough reverse-engineered set of data from the graph in the Economist (i.e. I looked at their graph and tried to guess their data points); I assumed that the Y axis on their graph reflects an absolute increase in home prices, although a percentage increase is just as likely, but again, they don’t give us any units so we have no way of knowing for sure. The yellow line shows the actual (reverse-engineered) performance of the Japanese market, while the other non-black lines show the same performance time-shifted to overlap with different sections of the US data; in all cases, the Japanese data is also shifted so that its initial value is equal to the value of the US market for the quarter in question.

The Economist would have us believe that you can predict crashes in markets by similarities in trend lines between different markets at different times; since our current trend is similar to the trend of the Japanese market just before that market crashed, they would have you believe, our market is also likely to crash. Note, however, that on my graph the purple shifted-Japanese-market line starting in 1990 matches our actual market behaviour quite well; using their logic, our housing market should have crashed catastrophically starting in 2001. Even the green line starting in 1985 matches ours fairly well (as do the ones for the intervening years, which I omitted for readability), but no decades long downturn occurred here in 1996 (as the green line would have predicted)—or any other year, for that matter. If the same superimposition could be made in multiple places on a graph, then it’s incumbent on the graph-maker to explain why the superimposition was made in the particular place that it was; otherwise, our only reasonable conclusion is that they chose to compare the two periods that they did for no reason but simple demagoguery.

You can shift the Japanese housing market data around however you want to compare it to different snapshots of the US housing market, but the simple fact is that the US economy is very different from the Japanese economy, and trying to predict the future using a single graph based on dubious data (I just can’t point out enough how little credibility I give someone who says “the value in 1985 was 100″ right below a graph of their own making that claims that that value was 140) is a sign of either ineptitude or a desire to intentionally mislead. That’s not to say that our market doesn’t bear any resemblance to the Japanese market; there are, in fact, some similarities. The biggest factor affecting Japan’s economy during the majority of their housing downturn, though, was (and continues to be) deflation; the fact that the US economy isn’t experiencing a deflationary period right now means that, by definition, any comparison of the two markets is an apples-to-oranges comparison that requires a lot more discussion of how and why the values in question differ.

(NOTE: after creating that graph, I noticed that the program I wrote to generate it accidentally extended the tails on the Japanese data, so each line extends for 5 years longer than the legend claims it does. Unlike the folks at the Economist, however, I actually noticed that mistake, and if that detail actually affected the interpretation of the data (or if I were actually trying to interpret the data at all, rather than simply point out how easy it is to manipulate and/or mis-interpret the data by formatting it misleadingly), I would fix it.)

No Reasonable Explanations

Going beyond their slanted presentation of information, the article also doesn’t provide any strong explanation for what they imply that they see in the statistics.

For one thing, they never define “bubble”; there is an implication in what they say that the rise in the ratio of total value of residential property vs. GDP over 5 years is the definition, but that seems quite arbitrary. That’s not to say it’s irrelevant, but without a rigorous justification, it can’t be used to explain anything with any credibility. Why not? Taking an example from the world of stock bubbles: financial writer Jason Zweig, in a 1999 article in Money magazine and later in his commentary on the investment classic The Intelligent Investor (written by Ben Graham, the inventor of “value investing” and the mentor to Warren Buffett), describes how in the 1990s the investment advice group The Motley Fool promoted an investment theory called “The Foolish Four”, which they showed would have historically beaten the market by more than 10% annually, and which was based on some simple rules and mathematical criteria that at first glance seemed very reasonable. As Zweig points out, however, upon a more rigorous examination, those rules (which essentially involve picking the 5 “best values” out of the 30 stocks that compose the Dow Jones Industrial Average, and then investing in the four of those five with the highest price) actually make very little sense—why, for example, would it be wise to not consider buying the stock of the cheapest-priced “value” stock, especially given that one of the other rules used says that investors should choose the cheapest-priced stock when there is a tie for which stocks fall in that “5 best values” category? Foolish Four proponents gave answers that again appeared at first blush to make sense—and then, in the December 2000, the originators of the Foolish Four did a more rigorous examination of the math underlying their theory, and themselves came to the conclusion that after taxes it actually underperformed the market, and stopped recommending it as a general purpose investment theory. Zweig then goes on to point out how arbitrary statistical correlations can be used to support almost any market prediction theory; after analyzing the performance of more than 10,000 stocks over a 20 year period, for example, he “shows” how a portfolio composed only of stocks without repeated letters (as he puts it: “Texaco, good… Exxon, bad”) would (using the same sort of logic that The Economist does) be expected to beat the general market.

Using an arbitrary ratio isn’t The Economist’s only sin against good analysis. By juxtaposing the current value of their market cap vs. GDP ratio worldwide against the same ratio before the American stock market crashes in the 1990s and 1920s, they imply that there is a similar pending crash in the world housing market. The world housing market, however, is not a stock market, and in particular is not the US stock market. US stocks are fungible, highly liquid, usually bought without any leverage, and of little practical use beyond appreciation and income; houses, on the other hand, are unique, highly illiquid, usually bought on margin, and (for owner-occupied properties) highly functional even if their value declines (because you can always continue to live in them, assuming that the circumstances that caused a value decline (such as a recession) don’t also cause a drop in your ability to pay your mortgage). The dynamics of a housing market are thus very different from those of a stock market; an implied correlation between a ratio (whose relevance, again, has never been explained) in one market and that same ratio in such a fundamentally different market is not what one should be basing a thesis on.

It should be obvious by now that I don’t like how The Economist presented this issue. Most people, sadly, probably don’t care about the presentation; to them, the only thing of relevance is: do I think that there is a housing bubble? I would direct those people to this article that came across the AP wire a few years ago about changes caused in frogs that were exposed to the weed-killer atrazine, and in particular to the response from Dr. Tyrone Hayes when reporters (who obviously would have preferred to write an article saying that atrazine was harmful to humans) pressed him about the chemical’s effect on people:

I’m not saying it’s safe for humans. I’m not saying it’s unsafe for humans. All I’m saying is it that it makes hermaphrodites of frogs.

I intentionally used that same language in my introduction, because that’s the only real answer that an intelligent and honest analyst can give until they have enough information to make a full and informed assessment: “here are the facts that I currently have, and I’m not willing to reach (or imply) any conclusions until I’ve looked at the problem further”. My own opinion is that the local Metro DC market is inflated, but with underlying fundamentals that will support that inflation and prevent any significant crash; that opinion is based on a lot of data and research into how markets function, but without more rigorous analysis of that data, the only thing I’m ready to say with 100% certainty right now is that housing bubble stories make frogs of The Economist.