By Ben S. Bernanke
Why was the Great Recession so deep? Certainly, the collapse of the housing bubble was the key precipitating event; falling house prices depressed consumer wealth and spending while leading to sharp reductions in residential construction. However, as I argue in a new paper and blog post, the most damaging aspect of the unwinding bubble was that it ultimately touched off a broad-based financial panic, including runs on wholesale funding and indiscriminate fire sales of even non-mortgage credit. The panic in turn choked off credit supply, pushing the economy into a much more severe decline than otherwise would have occurred. My evidence for this claim is that indicators of panic, including the sharp increases in funding costs for financial institutions and the spiking yields on securitized non-mortgage assets, are strikingly better predictors of the timing and depth of the recession than are housing-related variables such as house prices, market pricing of subprime mortgages, or mortgage delinquency rates.
In a recent post, Paul Krugman gave his take on the causes of the Great Recession. His inclination, contrary to my findings, is to emphasize the effects of the housing bust on aggregate demand rather than the financial panic as the source of the downturn. In a follow-up response to my paper, Krugman asks for evidence on the transmission mechanism. Specifically, if the financial disruption was the major cause of the recession, how were its effects reflected in the major components of GDP, such as consumption and investment? In this post I’ll offer a few thoughts on Paul’s questions.
I’ll start with some observations on the transmission mechanism. Certainly, a reduction in credit supply will affect normally credit-sensitive components of spending, like capital investment, as Krugman notes. But a broad-based and violent financial panic, like the one that gripped the country a decade ago, will also affect the behavior of even firms and households not currently seeking new loans. For example, in a panic, any firm that relies on credit to finance its ongoing operations (such as major corporations that rely on commercial paper) or that might need credit in the near future will face strong incentives to conserve cash and increase precautionary savings. For many firms, the fastest way to cut costs is to lay off workers, rather than to hoard labor and build inventories in the face of slowing demand, as they might normally do. That appears to be what happened: Job losses, which averaged 120,000 per month from the beginning of the recession in December 2007 through August 2008, accelerated to 670,000 per month from September 2008 through March 2009, the period of most intense panic. The unemployment rate, which—despite the fact that house prices had been falling for more than two years — was still around 6 percent in September 2008, shot up almost 4 percentage points over the next year. These are not small effects. Workers, in turn, having been laid off or knowing that they might be, and expecting a lack of access to credit, would likewise have had every incentive to reduce spending and to try to build up cash buffers. Indeed, research has found significant increases in precautionary savings during the financial crisis for both households and firms. In Krugman’s preferred IS-LM terminology, the panic induced a large downward shift in the IS curve.
Although isolating the effects of the credit shock on individual spending components is difficult, it’s nevertheless interesting to follow Krugman and examine how key components of GDP behaved during the recession. The chart below shows real residential investment and real GDP (all data below are quarterly, at annualized growth rates) for the period 2006-2009. As Krugman points out, there were large declines in residential investment in 2006-2007, prior to the major disruptions in financial markets. That’s consistent with his “housing bust” theory of the recession. However, note two points. First, despite the decline in residential investment in 2006-07, real GDP growth remained positive until the first quarter of 2008 and declined only very slightly over the first three quarters of that year, giving little hint of what was to come. However, after the crisis intensified in August/September 2008, GDP fell at annual rates of 8.4 percent in the fourth quarter of 2008 and 4.4 percent in the first quarter of 2009. That precipitous decline ended and began to reverse only as the panic was controlled in the spring of 2009.
Second, the pattern of residential investment was itself evidently affected by the panic, accelerating its pace of decline to a remarkable -34 percent at an annual rate in the fourth quarter of 2008 and -33 percent in the first quarter of 2009, before stabilizing in the second half of 2009 as the panic subsided. That the panic would affect the pace of homebuilding makes intuitive sense, given the reliance on credit of both construction companies and homebuyers. Indeed, my research finds that housing-related indicators like house prices and subprime mortgage valuations predict housing starts reasonably well through 2007, but that after that, indicators of financial panic, including the yields on non-mortgage credit, are actually better predictors of housing activity. In short, absent the panic, the pace and extent of the decline in the housing sector might itself not have been as severe.
The next chart shows the growth of nonresidential fixed business investment, whose behavior Krugman also cites in favor of the housing bust view. But here again, the timing is key to the interpretation. Unlike residential investment, which began contracting early in 2006, business investment did not start to decline until well after the bursting of the housing bubble. From the start of 2006 through the third quarter of 2007, as house prices fell, nonresidential fixed investment growth averaged almost 8 percent, in line with or even above pre-crisis norms. From the beginning of the recession in the fourth quarter of 2007 to the third quarter of 2008, average investment growth was slow but positive. However, from the fourth quarter of 2008, when the panic became intense, through the end of the recession in mid-2009, the rate of business investment growth fell precipitously, to an average annualized rate of -20 percent. Essentially all the decline of business investment took place during the period of most intense panic.
The next two charts show the growth of (1) real personal consumption expenditures for durable goods and (2) the components of the US trade balance. As with business investment, the worst declines in these series took place during the period of extreme panic. In particular, consumer durables spending remained healthy throughout 2006 and 2007, despite declining house prices and home construction. However, in the fourth quarter of 2008, durables spending declined at a 26 percent annual rate, recovering in early 2009 as the panic ended. Likewise, over the fourth quarter of 2008 and the first quarter of 2009, real exports and real imports both fell at average annualized rates of close to 24 percent, as global trade contracted sharply.
Because both exports and imports fell, the net contribution of trade to U.S. aggregate demand was modest. The behavior of the components of trade shown in the figure is nevertheless interesting for this discussion. Trade is particularly credit-sensitive, because importers and exporters rely on trade finance and because a significant portion of trade is in durables, a credit-sensitive category. The collapse of trade in late 2008 and early 2009 is therefore a reasonably good signal of disruptions in credit supply. Likewise, improvements in trade in 2009 likely reflected policies that ended the panic. Expanding on the international theme, note also that the global financial crisis can explain, in a way that the U.S. housing bubble cannot, the depth and synchronization of the worldwide recession of 2008-2009. (See for example, recent analysis by the Bank of England.)
To be clear, none of this disputes that the housing bubble and its unwinding was an essential cause of the recession. Besides their direct effects on demand, the problems in housing and mortgage markets provided the spark that ignited the panic; and the slow recovery from the initial downturn likely was due in part to deleveraging by households and firms exposed to the housing sector. Indeed, my own past research argues that factors related to balance sheet deleveraging and the so-called financial accelerator can have important effects on the pace of economic growth. I do claim, though, that if the financial system had been strong enough to absorb the collapse of the housing bubble without falling into panic, the Great Recession would have been significantly less great. By the same token, if the panic had not been contained by a forceful government response, the economic costs would have been much greater.
One more piece of evidence on this point comes from contemporaneous macroeconomic forecasts. Forecasts made in 2008, by both government agencies and private forecasters, typically incorporated severe declines in house prices and construction among their assumptions but still did not anticipate the severity of the downturn. For example, as discussed in a recent paper by Don Kohn and Brian Sack, the Fed staff’s August 2008 Greenbook report included economic forecasts under a “severe financial stress scenario.” Among the assumptions of this conditional forecast were that house prices would decline by an additional 10 percent relative to baseline forecasts (which had already incorporated significant declines). As a result, the assumed declines in house prices in this projection were close to those that actually would occur. However, even with these assumptions, Fed economists predicted that the unemployment rate would peak at only 6.7 percent, compared to its actual peak of around 10 percent in the fall of 2009. This conditional forecast would have taken full account of a sharp expected decline in housing construction and the wealth effects of falling house prices. The fact that forecasts still badly underestimated the rise in unemployment and the depth of the downturn suggests that some other factors—the financial panic, in my view — would play an important role in the contraction.
The failure of conventional economic models to forecast the effects of the financial panic relates to another point made by Krugman in a more recent post, in which he argues that the experience of the crisis and the Great Recession validates traditional macroeconomics. On many counts—such as the prediction that the Fed’s monetary policies would not be inflationary — I did and still do agree with him. However, as I discuss in my paper, current macro models still do not adequately account for the effects of credit-market conditions or financial instability on real activity. It’s an area where much more work is needed.