2008 was a landmark year in financial history. Two major financial institutions — Washington Mutual and Lehman Brothers — failed. Other major institutions such as the insurer AIG and the mortgage institutions Federal National Mortgage Association (Fannie) and the Federal Home Loan Mortgage Corporation (Freddie) survived only because of government interference. By autumn, interbank lending and commercial paper markets could not function normally because some of the participants were thought to have failure risks. The malfunctioning of these markets made near-term failures even more likely, which presumably aggravated the malfunctioning.
For good reason, there is concern that these financial events will damage the wider economy. Predicting the extent of wider damage requires an understanding of the causes of the financial turmoil. The purpose of this essay is to articulate a couple of hypotheses as to the fundamental origins of the crisis — the tastes, technologies, market structures, and public policies that may have caused these events. I particularly focus on causes of the cycle in housing prices and construction.
All of the hypotheses have a common chain of logic: something caused a massive housing boom between the years 2000-2006. The housing boom ended, requiring seismic adjustments by financial institutions, investors, homeowners, and the housing industry. In some cases, that “something” is tastes and technologies, which means that public policy should not and cannot prevent these events, but can at most affect their incidence and marginally change their magnitude. In other cases, events were caused or exacerbated by innate market weaknesses, which opens the possibility that public policy might improve matters.
Housing cycle theories differ in a couple of ways. Some of them focus on changes in tastes, technologies, and public policies during the boom years. Others focus on the expectations of market participants during those years about events that would occur in the future; those events may pertain to tastes or technologies. Among the expectation theories, some of them are “rational” and others feature bubbles or “irrational exuberance.” I explain how expectations about tastes differ from expectations about technologies in terms of how the housing bust might affect the economy going forward.
Analytical Framework and the Role of Anticipation
A variety of measures show real housing prices much higher at their peak in 2006 than they were in 2000 or during much of the 1990s. The OFHEO index shows prices grew 30 percent more 2000-2006 than the PPI for housing construction. The Case-Shiller index shows 90 percent growth in housing prices relative to PPI over that period. [1] Moreover, year 2000 real prices were somewhat more than they were during much of the 1990s. Housing construction activity followed a time pattern like that for prices: increasing through 2006, and then dropping steeply thereafter. Why did housing prices appreciate so much? Why did they fall so quickly?
A physical housing structure has value because it helps provide people with flows of shelter, privacy, and convenience. However, landlords and homeowners know that housing services flows are not produced with structures alone: also important are intermediate inputs of brokerage, management and banking services that can match and maintain the physical structures with the families who live in them and the investors who build them. As shown in the national accounts, banking, real estate brokerages, and others outside the construction industry annually supply hundreds of billions of dollars worth of the intermediate inputs to the housing industry.
Housing occupants pay rent for access to housing services. In the case of a tenant-landlord relationship, the rental transaction is explicit: the renter writes a monthly check to the landlord. In the case of owner-occupied housing, the rental transaction is implicit but economically just as real. The important point here is that the entire rent paid by occupants does not go to the investor in the structure — much of it goes to the suppliers of the intermediate inputs. The purchase price of a house (that is, the structure itself) is the expected present value of just that part of the rent that is left over after intermediate inputs are paid. The demand for residential structures is a derived demand; it depends not only on the demand for housing services but also the costs of intermediate inputs. The rental cost of housing services exceeds the capital rental received by the owner of a residential structure by the amount of intermediate input expenditure.
It is easy to show that the housing boom was possible only because of anticipated changes in housing service demand or in the costs of housing’s intermediate inputs. Specifically, the CPI for housing rent was only about four percent higher relative to the overall CPI during the years 2002-2006, as compared to its value in 2000. If the fraction of housing rent going to the owner of the structure were constant, and real housing rent was expected to remain four percent higher for the indefinite future, at most it would rationalize only four percent higher prices for the structures. The CPI for housing rent may be imperfect, but it seems clear that the housing boom involved dramatic increases in housing prices relative to housing rent. [2] Moreover, the intermediate inputs compensated by housing rent were pretty constant over these years (Mulligan, 2008b). Thus, home buyers during the boom years and/or their lenders must have anticipated that (a) housing rents would someday increase (or that a set of new buyers would someday emerge who expect housing rents to increase) or (b) an increasing fraction of housing rents would go to owners of the residential structures rather than suppliers of intermediate inputs.
Some optimism about the costs of intermediate inputs was warranted because applied information technology was rapidly advancing during the 2000s. Much of the banking and real estate value added relates to information. Bankers screen borrowers and value heterogeneous collateral. Real estate brokers match heterogeneous families to heterogeneous properties. Hall and Woodward (2008) claim
Recent years have seen great improvements in data, especially the introduction of credit scores, which gave lenders new powers to forecast mortgage defaults and to adjust interest rates offered to prospective borrowers. In 1990, credit scores were rare; by 1996, they were standard.
Perhaps lenders also expected to use information technology to better monitor and collect on loans, and therefore put subprime lending programs in place. Real estate brokers might also expect to benefit from technical progress: Virtual Office Web sites and their technological descendants might significantly reduce the brokerage resources needed to match homes with the persons who value them most.
Perhaps market participants hoped that information technological advance would eventually reduce the fraction of housing rent going to banks and brokers (increase the fraction going to owners of residential structures), and thereby support a greater equilibrium ratio of structures’ prices to housing rents. In this view, these expected technological advances (i) increased the equilibrium expected long run housing stock, (ii) increased short run housing prices and housing construction, (iii) increased short run expenditures on the intermediate inputs, and (iv) reduced the equilibrium expected long run housing service rent. [3]
Distinctions between Taste Change, Technological Change, and Bubbles
The observation that housing prices 2000-2006 were high and rising is sometimes taken as evidence of a rational bubble or of “irrational exuberance.” However, this observation is also consistent with a standard q-theory model in which expectations are rational, investors understood that real housing prices would eventually return to construction cost (and therefore to the levels of the late 1990s), and transversality conditions hold. Specifically, future technological advances (or future demand shifts — see below) create (v) anticipated capital gains for structures owners for the time prior to the ultimate increase in the fraction of housing rents going to structures owners. When the news arrives that the future technology will be better, structures prices immediately jump up but nothing improves the rental income received by a given structure in the short run. Those structures therefore earn less rent per dollar of capital for the time prior to the technological advance. Owners of structures rationally anticipate steady capital gains that compensate them over that period for the low rent per dollar invested. [4] If the technology expectations had been fully realized, housing prices would have peaked when the technological advances were complete, after which time housing prices would slowly fall back to construction cost while the housing stock rose to its long run amount and expenditures on the intermediate inputs were lower.
As with any anticipation, the ultimate results may be better or worse than expected. In this case, the technological advances were not fully forthcoming, or not forthcoming as rapidly as once anticipated. The bad news brings down housing prices. If the news were bad enough that the revised long run housing stock were at or below the actual stock at the time the bad news arrived, then housing prices would fall at least to construction cost.
A related hypothesis is that the housing boom was caused by the anticipation of high future housing demand, and that the crash occurred when it was learned that the high demand would never materialize, or would materialize later than was originally anticipated. Anticipated taste changes have some of the same effects as anticipated technological progress, because the taste for housing services and the costs of intermediate inputs both can cause shifts of the derived demand for residential structures. The implications (i) – (iii) for pricing, short run housing construction, and short run intermediate input use are therefore much the same. For the same reason, the expectation of high demand in the future creates anticipated capital gains for structures owners for the time prior to the ultimate increase in housing service rents, even though market participants understand that housing prices will eventually fall back to construction cost.
Anticipated taste and technical change are different in terms of expected future housing market outcomes and different in terms of how current behavior responds to those expectations. If and when the time ultimately comes when housing service demand is high, housing service rents are higher until the housing stock fully adjusts. Intermediate input use is higher than it was before the demand increase, merely because of the greater housing stock. As a result, fewer resources are available for the nonresidential sector. In contrast, the arrival of a better banking or brokerage technology ultimately reduces housing service rents as the stock of housing accumulates. Aggregate expenditure on intermediate inputs may ultimately be lower because of the technical progress — leaving more resources available for the nonresidential sector.
Because both the taste and technology hypotheses explain the housing bust as unfulfilled expectations, we cannot directly measure the equilibrium amounts of housing service rent and intermediate input use that would have occurred had the optimistic expectations been fulfilled. However, the hypotheses differ in terms of how market participants would have behaved during the boom and after the bust because they differ about the effect of the housing events on the amount of resources that would be available for the nonresidential sector. In other words, anticipated technological progress creates both a housing price increase and an aggregate wealth effect, whereas anticipated high demand for housing services may only increase housing prices without an aggregate wealth effect. [5] Bad news about future technological progress creates both a housing price crash and an adverse aggregate wealth effect, whereas anticipated high demand for housing services may only reduce housing prices without an aggregate wealth effect.
Quantitative empirical work more easily rejects taste and technology theories than theories of bubbles and irrational exuberance. Rational expectations theory limits how high housing prices can go, based on limits on population growth, housing budget shares, the time until tastes and technologies are expected to change, and the fact that intermediate inputs cannot be negative or be more complementary with structures than fixed proportions. I have not yet seen quantitative work on the amount housing prices should have increased based on fundamentals. [6]
Easy Money, Lax Regulation, and Other Possible Mortgage Distortions
Another hypothesis is that housing was excessively subsidized for a period of time roughly coinciding with the housing boom, and that the housing price reductions occurred as the end of the subsidy came near. In one version of this view, housing prices were high because the rental income from housing services — after taxes and subsidies — was higher than normal during the boom. This view is quite different from the expectations models sketched above, because in theory the (properly measured) rental income received by the owners of structures would be high during the housing boom, even when measured as a ratio to housing prices. As a result, owners of structures anticipate capital losses — even while housing construction booms — that compensate them for the high short-run ratio of rental income to structures price. To the extent that the subsidies served to reduce expenditures on intermediate inputs or were concentrated in the owner-occupied sector, housing rents paid by tenants to landlords per unit of housing services should have been low during the boom because rental housing has to compete with owner-occupied housing.
If this version of the subsidy hypothesis has any truth to it, it must be outweighed by the taste, technology, or bubble hypotheses because it is qualitatively inconsistent with two basic facts. First, housing rental rates were not low when measured per unit of housing services. Second, structures owners received continual capital gains, not capital losses, throughout the period of rapid housing construction.
Another version of the subsidy hypothesis says that public policy encouraged low mortgage rates, which raised housing prices. I believe that housing prices would not have gotten so high if mortgage rates had been higher, but low mortgage rates may not explain why 2006 housing prices were so high relative to housing prices in 2003 or 2008. 30-year fixed-mortgage rates were around six percent per year for most of the boom, and continue to be about six percent. The housing boom witnessed the increased availability of new mortgage products, such as interest-only loans. However, more work is needed to determine the degree to which the products reflected high prices, rather than product availability raising prices.
Conclusions
The methodology of economics specifies a clear recipe for assigning blame for today’s financial crisis. First, alternative hypotheses are articulated. Second, the alternatives are compared on the basis of empirical predictions. Third, empirical tests are conducted to gauge the relative importance of the alternative hypotheses. Unfortunately, October and November of 2008 witnessed much confident commentary as to the causes of the crisis, without much open, public application of the economic methodology.
My work attempts to follow these steps as regards the housing cycle from 2000 to the present. I explain that the housing cycle cannot necessarily be blamed on (supposedly) innate weaknesses of the marketplace, because a well-functioning marketplace that anticipated eventual technical progress in mortgages and real estate brokerage and/or eventually high future housing demand would have produced high and rising prices of housing structures. If and when future technical progress or demand was learned to be less than expected, housing prices would have crashed even in a well-functioning market. Of course, these theoretical results do not prove that the housing market was functioning well. But nor do the housing market observations 2000-2008 necessarily reveal that the market has innate weaknesses, or that its innate weaknesses have by themselves created much social cost.
For similar reasons, it is difficult to know whether the high and rising housing prices prior to 2006 reflected optimism about tastes or technologies. However, measurement of the existence and direction of wealth effects on behavior can help gauge the relative importance of these two kinds of impulses. Bad news about future technological progress creates both a housing price crash and an adverse aggregate wealth effect, whereas anticipated high demand for housing services may only reduce housing prices without an aggregate wealth effect.
Space constraints limit the detail with which I can connect housing events to the financial turmoil, but a few basic results can be summarized. [7] First, housing construction and the value of existing nonfinancial businesses is expected to follow a qualitatively similar cycle — rising when housing prices rise and crashing when housing prices crash. Investment in the nonfinancial nonresidential sector is expected to follow the opposite pattern – low during the housing boom and high since the housing bust. Second, a variety of risk-sharing arrangements — especially collateralized home mortgages — serve to dissipate housing losses into the wider economy, with the banking sector as the conduit (Arrow, 2008). Banks’ costs of old mortgages are sunk, so there is no efficiency reason why massive losses by themselves would affect bank operations. However, financial institutions may merge and some of them lose market share as the financial market dissipates the housing losses and the housing market changes the types of financial services it demands. Third, the labor market may be harmed in the short run as lenders mean-test their forgiveness of prior mortgage debts (Mulligan 2008a). Thus, some of the economic and financial adjustments are largely unrelated to the fundamental cause of the housing cycle, and just stem from the fact that housing prices were once high and now are not.
Both optimal public policy and some other economic adjustments depend on the causes of the housing cycle. In particular, the aggregate wealth effects are different depending on whether the housing cycle was caused by tastes, technology, or public policy. If the housing bust was an adverse aggregate wealth effect, the economy will have to adjust to less consumption and more long-run employment. Finally, if housing prices are high and rising because of the advent of technological progress or increased demand, a mis-interpretation of events as “irrational exuberance” might rationalize public policies that retard housing supply, create housing shortages, and ultimately cause harm. Empirical testing of carefully articulated alternative hypotheses are needed to make the ultimate judgment. Until then, we honestly do not know whether market participants’ unregulated planning for the future was part of the problem, or should be an integral part of the solution.
—
Casey B. Mulligan is professor of economics at the University of Chicago.
Notes
I thank Kevin Murphy for first suggesting to me some of the possible roles of technical change. I also appreciate the comments and patience of Ed Glaeser, who experienced an even more burdensome exposition of these ideas. I will provide updates on this matter on my blog www.panic2008.net.
[1] The OFHEO index is considered by Glaeser, Gyourko, and Saiz, (2008). Mulligan and Threinen (2008) compare growth rates for the OFHEO and Case-Shiller indices 2000-2006.
[2] To the extent that owner-occupied housing is a different good from rental housing, the housing rent CPI could have understated the increase of implicit rental rates for owner-occupied housing during the housing boom.
[3] I assume that intermediate inputs and physical structures are complements in the production of housing services.
[4] This is a modified version of Poterba (1984) and Topel and Rosen (1988). For the phase diagrams for such a model, see Summers (1981).
[5] Buiter (2008) has a model in which housing booms do not have aggregate wealth effects, but his model has no role for the kinds of technological change I have discussed. An increase in the size of the economy’s production set has an aggregate wealth effect by any definition.
[6] One obstacle to such work is that the housing price in economic theory (the cost of a structure that would produce one unit of housing services) is different from the housing price measured by Case-Shiller or OFHEO or the price relevant to mortgages (the cost of a given property, including all of its improvements or depreciation).
[7] Some of the additional detail can be found in Mulligan and Threinen (2008).
References
Arrow, Kenneth. “Risky Business.” guardian.co.uk. October 15, 2008.
Buiter, Willem H. “Housing Wealth Isn’t Wealth.” NBER working paper no. 14204, July 2008.
Glaeser, Edward L., Joseph Gyourko, and Albert Saiz. “Housing Supply and Housing Bubbles.” Forthcoming, Journal of Urban Economics, 2008.
Hall, Robert E. and Susan E. Woodward. “The Financial Crisis and Recession.” Manuscript, Stanford University, November 24, 2008. (http://sites.google.com/site/woodwardhall/)
Mulligan, Casey B. “A Depressing Scenario: Mortgage Debt Becomes Unemployment Insurance.” NBER working paper no. 14514, November 2008a.
Mulligan, Casey B. “Housing Wealth and Aggregate Wealth Effects.” Manuscript, University of Chicago, November 2008b.
Mulligan, Casey B. and Luke Threinen. “Market Responses to the Panic of 2008.” NBER working paper no. 14446, October 2008.
Poterba, James M. “Tax Subsidies to Owner-Occupied Housing: An Asset Market Approach.” Quarterly Journal of Economics. 99, November 1984: 729-52.
Summers, Lawrence. “Taxation and Corporate Investment: A q Theory Approach.” Brookings Papers on Economic Activity. 1981(1): 67-127.
Topel, Robert and Sherwin Rosen. “Housing Investment in the United States.” Journal of Political Economy. 96(4), August 1988: 718-740.