This is a little technical, at least at first. Econ majors might be interested in all of it, everyone else can probably skim to the last four or five paragraphs (with which many econ majors will most likely be very uncomfortable since in them I discuss normative aspects of economic phenomena).
In a blog published last August, Harvard economist, Ed Glaeser, summarized the results of his econometrically heroic effort with two colleagues to examine the causal effect of low interest rates and other measurable attributes of "relaxed" credit markets, such as higher loan to value ratios, on housing prices during the recent bubble. He concludes in both blog and paper that "credit surely mattered, but I fear that it will take many years before we truly understand how."
It's frustrating when empirical analyses derived from economic theory do not fully account for observed phenomena. A lower "price," whether because lower interest rates have reduced both the monthly payment and the total amount borrowed or because looser borrowing standards require buyers to part with less money up front in the form of a down payment, should translate into increased demand. Increased demand for houses, all else held equal, should drive up the price of houses. Glaeser's study is complicated because he finds some evidence that reduced interest rates and lower down payments probably increased housing prices, but not by enough to account for the dramatic rise in prices between 1996 and 2006. It's further complicated by the fact that after 2006, even as interest rates fell, demand for housing did not rise as theory would predict (all else held equal).
One of the challenges that Glaeser and his colleagues confront is the dearth of good data. By that I mean data that are not measured with considerable error because a data source does not accurately capture key outcomes such as loan to value ratios or approval rates. Good data for such an analysis also requires that we have more detailed information about purchasers than race, sex, and income (the last measured with more error than usual in the case of "liar loans.") In an ideal world, the data we would like to have would include purchasers' expectations of future housing prices, the rates at which they discount future states of the world (their time preference), their levels of risk aversion, and the extent to which they understood the terms and conditions of the loan agreements they signed. All of these, in addition to interest rates and down payment amount, would be expected to influence individual demand for housing.
One key issue in assessing the extent to which home prices were driven by credit market characteristics is the extent to which characteristics of "marginal" purchasers (purchasers who would not have bought a house except in response to credit market changes) changed. If changes in credit availability and interest rates induced more risk takers (aka speculators) with expectations of higher increases in future housing prices to buy houses, I would expect this to drive housing prices higher than if the market were composed mostly of the prudently risk averse. Similarly, if prudently risk averse buyers and unqualified buyers were induced to buy houses they otherwise would not have bought by obscuring or disregarding risk completely, it would also tend to drive housing prices higher. Risk aversion on the part of buyers and accurately signalled risk in markets or by loan originators are important components of well-functioning markets, especially markets for assets, like houses, where values change over time in response to market conditions.
Then there are people who would have bought houses during the boom years no matter what, not in response to changes in interest rates or loan to value ratios, but simply because they moved or had been planning to trade up or down or to buy all along. I'm not saying that these buyers' decisions are not affected by price. My point is simply that in a counterfactual world without a housing bubble, some of the buyers who bought in the bubble would also have bought in the absence of a bubble. Glaeser et al's task if they are to establish a causal effect of interest rates and LTV ratios on housing prices is to identify the effect of interest and LTV ratios on those who bought houses in response to these two variables, to identify econometrically the purchasers who in a counterfactual world of higher interest and lending standards, i.e., in the absence of low interest rates and low down payments, would not have bought houses.
Add to the data problems the possibility of "reverse causality," that higher prices induced bankers and mortgage originators to relax lending standards because they could earn more from origination and other fees and, in a worst case scenario (they thought), could foreclose on and resell an asset that had gained in value. It becomes very difficult to tease out the true causal impact of low interest and lower down payments on housing prices because they are correlated with unmeasured buyer and banker traits and behavior that also influenced demand and prices.
Can you see why it's not easy being an economist who does applied research? Not only do you get no respect from the theoreticians, you spend weeks and months trying to figure out how to get credible estimates from data that were (more often than not) never designed to provide them. But I digress.
After reading Glaeser et al's paper, I found myself impressed by the thought and the econometric techniques he and his co-authors brought to bear in an effort to to answer their seemingly simple question with much less than perfect data:
Did the proliferation of low-down-payment, low-documentation loans cause the housing bubble?
One of the best things about the economics profession (IMHO) is that we have an impressive econometric toolkit for asking simple, but often hard to answer questions of complex non-experimental data. One of the worst things about our profession is that some of us (I'm sure I'm guilty) sometimes delude ourselves that sound methods can wholly compensate for inadequate data, the cost of delusion when we're wrong being lives, jobs, and less than ideal monetary and fiscal policies. For this reason, it was with an incredible sense of suspense that I read the Glaeser et al paper, fearing that at any moment they would conclude that their econometric heroics had produced credible results. I felt immense relief when they concluded that their results were probably not reliable except to demonstrate that, using sound technique, the answer seemed to be: not very much.
Why did I feel such suspense, I asked myself? I think it was mainly that this was one of many cases where sparse data are a real liability, mainly because bubbles are fueled by sentiment, beliefs, and attitudes. The stuff psychologists measure and worry about. Animal spirits, we call them. It's not surprising that empirical tests of economic theory break down in analyzing situations where animal spirits play important if not dominant roles. It's not that we wouldn't measure and include them if we could (we'd call them "preferences"). It's that the data sets we tend to use seldom measure such things. Behavioral economics will change this, but it will take time.
Animal spirits are in the error term of almost every econometric analysis that involves consumer behavior. In most analyses, they don't matter much. In frothy, bubbly markets, they do. Age, gender, race, and an often inaccurate measure of income are usually insufficient to capture animal spirits, the sentiments that cause you to roll the die of your life savings in hopes that you make some money when you sell the house in six months or three years. Reliable instruments for animal spirits are often few and far between. The problem is that the kinds of animal spirits that fuel a real estate or any other kind of bubble derive not just from individual risk preferences, they also derive from a rich, inter-related tapestry of beliefs about risk and future prices that comes to characterize many if not all participants in a bubble market. My beliefs become correlated with your beliefs and we all ride to hell in a hand basket together.
But maybe Glaeser et al asked the wrong question. Maybe there's a another question that is more relevant: what other features of the economy might combine with low interest rates and low credit standards to increase the marginal home buyer's desire to take chances or to ignore risk? Specifically, what is the effect of an economy in which low and middle incomes have been stagnating for nigh onto 30 years? Surely, some of those who bought homes induced by lower barriers and prices were simply trying to grab the gold ring that has been receding from their grasp for at least a generation. Salaries may not have risen, but there was apparently money to be made by leveraging what little they had into a home that historically had provided a source of wealth for many working and middle class families.
The allure of a casino is difficult to resist when other means for achieving economic security are diminished. It seems to me there's a more important question here than "did the proliferation of low-down-payment, low-documentation loans cause the housing bubble?" What I find myself wondering is: did the failure of our politics and our economy to deliver shared universal opulence over the last 30 years create incentives for many individuals to opt for speculation and imprudent participation in housing markets, instead of pursuing the (now non-existent) gains that used to accrue to productivity, prudent risk-taking, and creative innovation? If the answer to the last question is "yes," then we can hardly fault individuals for responding rationally to prevailing incentives.
Economists may not always get things right or agree on everything, but one thing I'm pretty sure we all agree on and that we are right about is that incentives matter. If we as a country want productivity, prudent risk-taking, and creative (and productive) innovation, we had best start rewarding it.