Why Housing?

Abstract

Asset bubbles come and go. Only the housing bubble, however, brought the economy to its knees. Why? What makes housing uniquely a cause of macroeconomic risk? This article examines the workings of the housing market as well as theories and empirical evidence about the housing bubble. It explains why housing is a particular source of macroeconomic risk and how changes in the housing finance channel were the critical element in the formation of the bubble.

Keywords:

• housing
• mortgages
• securitization

Notes

^1. There are other goods that are both consumption and investment goods, such as art, antiques, and jewelry, but they differ from housing in important ways. These other types of goods are not basic consumption goods. They are instead luxury goods that make up a small share of household balance sheets. And when consumers make bad investment decisions with these other types of goods, there are limited externalities, and people do not end up homeless.

^2. Home equity is determined by housing prices and the amount of borrowing, and changes in the availability of credit directly affect equity and household wealth. Greater availability of credit overcomes credit constraints and adds to demand for housing (Linneman & Wachter, 1989).

^3. Evicted households demand less housing, for example, through doubling up.

^4. Commercial leases are often assumed after foreclosures. Moreover, some commercial leases have subordination, nondisturbance, or attornment clauses that protect commercial real estate tenants in the event of a foreclosure of a mortgage entered into prior to their lease, but this is not universal. Therefore, it is possible that following a foreclosure, a tenant—even one current on the lease—can be evicted prior to expiration of the lease, but this is quite different from the owner-occupancy case. In owner-occupancy situations, a foreclosure requires the eviction of the current occupant (see Levitin & Wachter, 2013a).

^5. As discussed below, the linkages are through a balance sheet channel, a construction channel, and a financial sector collateral channel.

^6. Futures markets could also work, but they would need to be standard and liquid. Case/Shiller indexes trade but are not liquid. For further discussion, see Levitin, Pavlou, and Wachter (2012). CDSs could also work, but they are potentially subject to information problems, as we discuss below.

^7. Fostel and Geanakoplos (2011) show that with tranching the price of the underlying collateral can rise above what any agent thinks it is worth, while with leverage the price of the asset typically rises to what a more optimistic agent thinks it is worth.

^8. In other work, we lay out in detail the role information problems played in the bubble and how agency conflicts in mortgage finance enabled the exploitation of these information problems (Levitin & Wachter, 2012; Levitin et al., 2012).

^9. Elsewhere housing bubbles have formed through bank lending rather than securitization. Bank lending, which is itself opaque and subject to information problems, and financing of real estate bubbles is implicated in the recent bubble and crash in Ireland, Spain, and Iceland, as well as the Asian Financial Crisis and the earlier Japan real estate bubble and crash (Green, Mariano, Pavlov, & Wachter, 2009; Herring & Wacher, 1999). Thus, this is not to say that the housing bubbles cannot occur in the absence of securitization; banks can finance housing bubbles as well (Levitin et al., 2012).

^10. Among these omitted variables are market-specific credit constraints and the cost of credit. Market rates are needed because they represent the price of the borrower's put option on a nonrecourse loan—the option to borrow cheaply (if underpriced) today and either book the gains or walk if prices fall. Underpriced credit also heightens market risk because the underpricing will be revealed in future states, leading to sudden stops in credit. An inherent instability in the finance being supplied results if interest rates were too low for the risky characteristics of the loans in the first place. Pavlov and Wachter (2011) used the share of subprime lending (and instrument for this) as a proxy for credit constraints and the price of risk, to predict MSA prices in the cross section in up and down markets, all else equal.

^11. Second liens matter because levels of LTV ratios in the market, and importantly the distribution of LTVs, matter. If the average LTV and consolidated LTV (including second liens) for the stock of leveraged homes are reasonable, but a large share of the newly issued mortgages at existing prices has high LTVs whereas others' are low, as was the case, the result is more unstable than if all LTVs are at the average.

^12. Mortgage rates alone do not suffice; credit terms and credit supply are needed as well. Stiglitz and Weiss (1981) showed that adverse selection and moral hazard lead to a role for credit rationing (rather than market clearance through interest rates).

^13. Foote et al. (2012) point to the active trading of PLS and specifically assert that there was rich information available about the pricing of individual mortgages and individual PLS and that information on the state of the credit market was in fact widely available. Foote et al. however, consider neither the aggregate of credit characteristics, which is at least potentially knowable but was not known, nor the state of credit relative to housing prices for markets across the country. A bubble cannot be identified from a sampling of loans, such as the information for particular deals, which is what Foote et al. note were available to investors in the 2000s. But to understand the impact of credit on housing prices, it is necessary to know the aggregate supply of credit and the attributes of that supply, as well as the distribution of that credit in housing markets, not merely anecdotal information on individual loans or samples. Aggregate data are necessary to model the impact of a potential change in credit and economic conditions on future credit availability and, therefore, on the performance of mortgages, including not only mortgages that require refinancing but also those mortgages potentially impacted by an increase in foreclosure rates.

^14. Mandating standardization of both mortgages and MBS to a limited number of options, such as the Neapolitan combination of plain vanilla, chocolate, and strawberry (but not Rocky Road), that we have proposed in prior work would be a starting point, but it would also require meaningful enforcement of soft underwriting qualities to be fully effective (Levitin & Wachter, 2012).

^15. Commercial investors will be diversified and this is diversifiable risk, so they should not be getting a premium for this risk. If there is a compression associated with a price rise in commercial real estate, then there is a bubble.

^16. A different motivation for assuming a risky loan would simply be that the borrower likes the odds of the gamble of the put option, given that mortgages generally are nonrecourse loans.

^17. A candidate for the technological innovation could have been automated underwriting. However, automated underwriting failed to predict performance of an aggregate book of business when terms shift, much as Lucas (1976) theorized.

^18. It is also possible that high price increases occurred because of the underpricing of the put option by lenders and investors. See footnote above.

^19. Without reliable short pressure, as noted, housing prices are set by optimists at the margin, creating fertile ground for Shiller's exuberant investors. Even a very short period of optimism, such as in 2004–2007, could in theory lead to a housing bubble, although given the importance of credit for home purchases, buyer beliefs alone are not likely to fuel a bubble, as confirmed by the literature, reviewed above. Moreover, a recent paper by Case and Shiller appears to draw doubt on expectations as a single explanatory factor. On the basis of a sample of surveys of homebuyers, Case, Shiller, and Thompson (2012) find that buyers' expectations confirm house price growth in the market but that their expectations do not get ahead of prices. Moreover, although buyers appear bullish on average, the high rate of appreciation that they expect begins to decline two years before prices peak in late 2006.

^20. The immediate impact of increased leverage is higher housing prices and a decrease in default risk.

^21. It is important to note that the financial technology involved in the shift in the financing channel was not new innovations. Private-label securitization, CDOs, and the OTD business model had all existed for decades, as Foote et al. note. What was new, however, was the emergence of these technologies from being niche products to becoming the market and the growth in complexity in the securitization market. The bubble was not the result of financial innovation per se but of the mass use of niche products. The OTD business model using PLS and CDOs to distribute mortgage credit risk to investors was the defining feature of the bubble.

^22. A longer run perspective on the start of the bubble focuses on credit expansion as a function of income inequality (Rajan, 2010). Notably from 1980 onward, household credit expanded far more than GNP. (White, 2007), and Income inequality may be the deeper explanation for both the increase in borrowing and policy changes that encouraged it. One of these policy shifts was (ironically) disallowing deduction of credit card payments. When these debts were consolidated into mortgage debt, it decreased the cost of credit for borrowers. Mortgage debt puts homes at risk because it is not possible for borrowers to discharge their debt under bankruptcy without losing their homes (Wachter, Warren, & Bachieva, 2005).

^23. Glaeser et al. (2008) show that 20% of the price rises between 1998 and 2006 can be explained by interest rate movements.

^24. We have found the same phenomenon to exist in the commercial real estate market, in which, excluding multifamily, it is an entirely private securitization market (Levitin & Wachter, 2013a). The first-loss position in commercial mortgage-backed securities (CMBS) was traditionally held by a small number of sophisticated B-piece buyers. Beginning in 2004, these B-piece buyers were outbid by CDOs. With the advent of the CDO in the CMBS B-piece market, underwriting standards declined precipitously, resulting in a bubble that closely tracks the housing bubble.

^25. CDS protection competed with other forms of credit insurance, such as monoline bond insurers. Monoline insurers found themselves in ruinous competition—essentially a rate war—with the CDS as they slashed their premiums to hang on to market share, which was necessary to provide continuous cash flows.

^26. Some CDS protection sellers, such as synthetic CDOs, were designed to eliminate counterparty risk via dedicated funding. For CDS protection written by synthetic CDOs, there was assumed to be almost no counterparty risk because the synthetic CDOs were prefunded with super-safe assets, some of which turned out to be AAA-rated PLS.

^27. Last year, for example, the Wall Street Journal reported, “the nation's 10 largest mortgage lenders denied 26.8% of loan applications in 2010, an increase from 23.5% in 2009” (Timiraos & Tamman, 2011, p. 2). According to the analysis, these restrictions have lasted longer following this recession than they had following previous postwar recessions . Moreover, the credit score for the denied borrowers exceeded previous credit scores for accepted borrowers.

^28. Regional clustering of foreclosures makes the effect more marked. Agarwal, Ambrose, Chomsisengphet, and Sanders (2012) found that a 1% increase in foreclosures in a given region increased the odds of default for surrounding homeowners by 2.9%. Concentrated foreclosures can produce lasting blight. Foreclosed properties fall into a state of disrepair when the owners leave, casting a pall on the neighborhood that depresses surrounding house prices.