We’ve all heard the story about the exceptionally large “housing wealth effect” that preceded the Great Recession. Homeowners felt increasingly rich as their homes boomed in value, and their newfound wealth encouraged them to spend like never before. “Using homes like they’re ATMs” was the cliché. Demand soared, production surged, and the economy boomed still further.
And then the bubble burst. When the housing market collapsed, so did spending. Supply exceeded demand; firms laid off workers. Now-jobless homeowners, saddled with mortgage debt that eclipsed their home’s declining value, simply stopped spending. The unparalleled wealth effect went into reverse and was even more potent on the way down, prolonging the recession.
Well, it looks like that story is wrong.
The housing wealth effect wasn’t particularly large in the 2000s, according to a new staff report from the Federal Reserve Bank of Minneapolis. In fact, the effect “was, if anything, smaller post-2000 than earlier,” write the authors. And it wasn’t larger on the downside.
The housing wealth effect wasn’t particularly large in the 2000s. The effect “was, if anything, smaller post-2000 than earlier.” And it wasn’t larger on the downside.
True, the housing sector played an outsized role in the economy, but simply because home prices soared so high, not because that housing balloon generated unprecedented spending. And the wealth effect wasn’t especially large or unique to the boom-and-bust of the 2000s. Rather, it’s a “fundamental feature of the economy going back to at least the 1980s.”
Taking a long view
In “Housing Wealth Effects: The Long View,” Minneapolis Fed senior research economist Alisdair McKay, with Adam Guren of Boston University, and Emi Nakamura and Jón Steinsson of the University of California, Berkeley, reaches these conclusions through careful examination of data on house price trends and economic activity, and close analysis of a detailed model of household behavior with constraints on borrowing, long-term housing debt, income risk, and illiquid housing markets.
They provide new estimates of “housing wealth elasticity”—their term for the sensitivity of economic activity to house prices—and those estimates offer no support for the notion that rising house prices of the early 2000s generated an exceptionally high level of consumer spending.
They also look into the oft-floated idea that spending is more sensitive to falling housing prices than rising, perhaps because during housing busts, homeowners find themselves “underwater,” owing more in mortgage debt than their homes are worth. There, too, the economists find “no statistically significant evidence of such a boom-bust asymmetry.”
New estimates provide no support for the notion that rising house prices of the early 2000s generated an exceptionally high level of consumer spending.
So, it seems, the Great Recession continues to burst bubbles: Long-held ideas about economic trends preceding and following the recession, and indeed at its very heart, have been deflated.
The study is exhaustive—a 96-page work by scholars known for scrupulous attention to statistical accuracy, analytical clarity, and thorough examination of alternative explanations and neglected variables.
Addressing the challenges
The economists begin by addressing a particular challenge in measuring housing wealth elasticity. Home prices and economic activity are “jointly determined”—higher economic activity includes more home buying, and more home buying is part of increased economic activity; causation can run in both directions. Another issue: measurement error of local housing prices.
The economists solve these and other statistical challenges to estimating housing wealth elasticity by using panel analysis—examining the same statistical sample over multiple time periods—and by creating a new instrument for city-level house price changes.
Their instrument isolates changes in local home prices that are driven by national home-price trends from those caused by changes in local economic activity. That enables them to rule out reverse causation and measure the degree to which local home price trends impact local activity—in other words: housing wealth elasticity.
To gauge consumer spending, they use retail employment data, the standard choice of econometricians because retail employment correlates closely with measures of aggregate consumption and, unlike other potential variables, relevant data are available over many decades.
They look at the housing wealth elasticity using their new instrument, and they analyze data for several decades, from the 1980s to 2017. For comparison, they analyze the same data using two other measures.
The main empirical finding: Historical trends in housing wealth elasticity are similar for all three methods. And for two of them, including their new instrument, elasticity was “significantly smaller” during the 2000-2012 boom-bust episode than during other years.
For the entire time span, back to the late 1980s, they calculate an elasticity of 0.072, which corresponds to a “marginal propensity to consume out of housing wealth of 3.3 cents on the dollar”—if local house prices rise $1,000, homeowners will likely spend an additional $33. It’s economically significant, but not massive. Definitely not an ATM.
A model analysis
Their next step is building a theoretical model to analyze mechanisms behind their empirical findings—a theory to explain the relative stability in housing wealth elasticity over time despite large changes during boom-bust cycles in both credit constraints and “loan-to-value” (LTV), the amount of mortgage debt held by homeowners relative to their home’s value.
Into a fairly standard model with borrowing constraints and income risk that can’t be insured, they incorporate features crucial for understanding housing wealth elasticity: long-term mortgages and illiquid housing (homes can’t be sold as quickly as, say, stocks or bonds). The model generates an average elasticity in line with their statistical estimate (0.09 model versus 0.072 data).
The model reveals two key forces that stabilize housing wealth elasticity over broad ranges of LTV.
Low loan-to-value households exhibit substantial and stable elasticity.
The first relates to households with low LTVs (mortgage debt equal to or below 60 percent of their home’s worth). This segment of the housing market exhibits substantial and very stable housing wealth elasticity. These low-LTV households constitute a large portion of the whole, about 38 percent of the U.S. population in 2007 and 62 percent of homeowners. “The overall housing wealth elasticity for this group is relatively stable in the face of shifts in the LTV distribution,” write the economists. “Since this is a large group, its stability contributes to stabilizing the aggregate housing wealth elasticity.”
The second factor is more complicated, relating to homeowner behavior when LTVs rise above 80 percent. For some of these households, spending becomes more sensitive to home value because it strongly affects the amount of home equity the household could access through home sale, refinancing, or home equity line of credit (HELOC).
As LTV rises above 80 percent, these households have less available equity to draw on. If their home values decline, they have to reduce spending. If their home values rise, they may feel more confident to raise spending since they now have a home equity buffer. For this group, housing wealth elasticity is therefore quite high.
But “once the LTV ratio rises above roughly 0.95,” the economists write, “the model-implied housing wealth elasticity drops rapidly.” Homeowners start to go underwater at that point and can no longer spend from their equity. They can’t refinance or sell, so instead they simply pay down their mortgages over time. “Their consumption is consequently highly insensitive to house price changes.”
These two effects offset one another: As LTV ratios rose in the Great Recession, some households became more sensitive to home prices, but others—those that went underwater—became less sensitive. Combining this two-effect factor with the first force of low-LTV households ultimately leads to stable housing wealth elasticity over a wide range of homeowner mortgage debt.
Still nothing special
Homeowners have exhibited the housing wealth effect for decades. In this regard, the 2000s boom-bust was not very different.
The model thus explains the empirical finding that the rise of consumer spending in response to home price appreciation in the 2000s and then its dual collapse was not exceptional and not driven by some special feature of the housing market during that traumatic business cycle. As the economists point out, HELOCs and other tools that homeowners can use to tap home equity have been around for years. They quote a report written by Alan Greenspan:
“Very rapidly rising prices for existing homes and a sharp increase in sales … of these homes has created a huge increase in capital gain and purchasing power.”
Greenspan wrote those words in 1982, well before he was chair of the Federal Reserve and decades before the Great Recession. In other words, homeowners have exhibited the housing wealth effect for decades. In this regard, the 2000s boom-bust was not very different.