The U.S. foreclosure crisis and the larger financial crisis it precipitated have had, and will continue to have, a wide variety of direct and indirect impacts on neighborhoods, cities, and metropolitan regions (Schilling, 2009). I begin by describing the development of the foreclosure crisis, its impacts, and the resulting accumulations of foreclosed properties among and within metropolitan areas. I then describe the federal policy response to the foreclosure and larger financial crisis up through mid-2009, focusing on efforts to reduce surging foreclosures and deal with the problems of vacant, foreclosed properties. I also discuss some potential implications of likely changes in housing finance for housing patterns and metropolitan development. My principal focus is on impacts that stem from likely changes in housing finance rather than from the broader economic downturn. This discussion is based largely on informed speculation, but also on some changes in housing finance that are already evident. Highly definitive predictions are certainly unwise at this point, since outcomes will be driven in large part by as yet unknown responses by policymakers and financial markets. However, the impacts of these changes, should they materialize, would likely be very large, so planning practitioners and scholars should start discussing and researching them sooner rather than later. Speculation around postcrisis changes in development patterns has clearly already begun (Florida, 2009; Leinberger, 2008).

Finally, I turn to discussions of more specific, and nearer- term, federal policy issues that have been stimulated or resurrected by the crisis and are likely to have substantial influence on housing and community development. I focus on issues of housing finance and neighborhood stabilization in the face of accumulating foreclosed properties, including: federal neighborhood stabilization funding, mortgage market regulation, the federal role in secondary markets and securitization, and community reinvestment and fair lending policies. The relevance of some of these topics to planning may not be immediately obvious, but the history of metropolitan development in the United States suggests that housing policy and especially housing finance can be key forces in shaping metropolitan regions. The next 10-20 years will likely be no exception.

The Evolution of High-Risk Mortgage Markets

Contrary to some media reports, the development of high-risk mortgage lending began well before the recent run up (Immergluck, 2009). An earlier boom in the second half of the 1990s was marked by a surge in subprime refinance lending. After 2001, subprime home purchase loans grew rapidly, together with a new class of exotic mortgages, alternative mortgages aimed at prime borrowers. Subprime home purchase loans grew over 250% from 2001 to 2004 (Immergluck, 2009). While subprime mortgages made during the first boom performed very poorly, the loans made during the second boom performed even worse. According to the Mortgage Bankers Association (2009), outstanding subprime loans were entering foreclosure at an annualized rate of over 17% by the second quarter of 2008. Moreover, this rate was based on a much larger population of subprime loans than in previous periods.

A key factor in the growth of the subprime market in the 1990s was the vertical disintegration of the lending industry as securitization grew and fewer originators held their own mortgage loans (Jacobides, 2005). However, mortgage securitization did not appear out of thin air. Deregulation and the federal preemption of state regulations laid the groundwork for increased securitization in the 1980s, initally by Fannie Mae and Freddie Mac, the governmentsponsored enterprises (GSEs) whose principal business was the bundling and securitization of mortgage loans (McCoy & Renuart, 2008). ' Even more importantly, deregulation and favorable tax and securities policies facilitated the growth of private-label securitization, which grew rapidly in the 1990s. Private-label mortgage-backed securities, which do not go through the GSEs, grew from $35 billion in 1993 to $150 billion in 1998 (U.S. Department of the Treasury & U.S. Department of Housing and Urban Development, 2000).

As dominance in the mortgage market shifted from savings and loans to mortgage companies in the 1980s and 1 990s, federal policymakers did little to adapt supervisory systems to the new market structure, effectively deregulating through lack of action (Immergluck, 2009). Congress passed the Home Ownership and Equity Protection Act (HOEPA) in 1994, giving the Federal Reserve Board the power to issue proscriptive regulations on both high-cost and non- high-cost loans. But the Board's protections primarily applied to the former category, which was defined by such high price thresholds that it covered few loans, and they had little impact. The statute gave the Federal Reserve the authority to add more proscriptive regulations and to lower thresholds, but it did very little in this regard until 2008, when it finally issued more regulations on a broader set of loans after subprime originators had essentially shut down.

Some states attempted to strengthen their own regulations in the late 1990s and early 2000s, but federal regulators, including the Office of the Comptroller of the Currency (OCC) and the Office of Thrift Supervision (OTS), maintained that their authorities took precedence over state laws for federally chartered lenders and their affiliates (McCoy & Renuart, 2008). This blocked the efforts of consumer advocates who had been fighting for stronger regulations at the state level.

High-risk lenders exploited the geographies of social disadvantage, and federal regulators failed to address geographical and racial disparities, even when given policy tools to do so (Apgar, Calder & Fauth, 2004; Squires, 2003; Wyly, Moos, Froxcroft, & Kabahizi, 2007). By 1998, subprime lenders dominated the refinance market in Black neighborhoods across the country. Subprime lenders made 51% of refinance loans in predominantly Black census tracts, compared to only 9% in predominantly White tracts (U.S. Department of the Treasury & U.S. Department of Housing and Urban Development, 2000). Refinance borrowers in upper-income Black tracts were six times more likely than borrowers in upper-income White tracts to receive subprime loans. Calem, Gillen, and Wachter (2004) found that, even after controlling for education, income, credit histories, and other characteristics, an allBlack tract was expected to have a subprime share that was 24 percentage points higher than an otherwise equivalent White tract.

Two new forces fed the second boom in high-risk lending. One was the rapid appreciation of home values in many metropolitan markets, especially in the West and Southwest, in Florida, and on some parts of the East Coast. Lenders responded to affordability problems in areas with rapidly escalating prices by developing new affordability products that offered exotic loan structures to both prime and subprime borrowers. As home prices rose, lenders increasingly competed for borrowers by offering larger loans, enabling the purchase of larger homes or homes in more desirable areas. Subprime and exotic lending increased the effective purchasing power of buyers in most markets, fueling price appreciation, which in turn led to more highrisk lending. In many places, greater purchasing power was largely transformed into higher home values (Green & Wachter, 2007).

Another factor contributing to the second high-risk lending boom was the increased supply of high-risk capital. As the dot-com bubble burst, many sellers chose to invest in real estate instead of stocks (Downs, 2007). What the Federal Reserve Chairman called a "global saving glut" also propelled capital into the United States. Net international lending to U.S. citizens, businesses, and governments increased from $120 billion in 1996 to $666 billion in 2004 (Bernanke, 2005).

Private-label securitization played an increasingly important role in fueling high-risk lending. Subprime and Alt-A^sup 2^ mortgage-backed securities increased from $98 billion in 2001 to approximately $814 billion by 2006 (Ashcraft & Schuermann, 2008). Financial innovation in securitization markets, especially the use of collateralized debt obligations (CDOs) and credit default swaps, increased the risk levels in mortgage markets. CDOs pool mortgagebacked security bonds, some with ratings below AAA, transforming lower grade mortgage-backed security bonds into higher- rated CDO bonds (Mason & Rosen, 2007). Credit default swaps are essentially private, unregulated, insurance agreements that allow investors in mortgagebacked securities and CDOs to hedge their investments, increasing the amount of capital flowing into such investments.

Securitization schemes created frictions between parties in the credit supply chain, including loan originators, credit rating agencies, issuers of securities, and investors (Ashcraft & Schuermann, 2008). These frictions often involved principal-agent or asymmetric information problems, and occurred when one party had an incentive to conceal critical information from another party. The complexity of mortgage-related securities made them less than transparent and caused investors to rely on ratings from the large credit rating agencies. These firms, including Standard & Poors, Moody's, and Fitch Ratings, repeatedly underestimated or understated the risks to investors in mortgage- backed securities and CDOs (Mason & Rosen, 2007). Finally, different tranches (groups of investors holding mortgagebacked securities, each group with a different maturity or rate of return) had different and sometimes conflicting interests if loan modifications were required (Eggert, 2007). This created a threat of litigation against loan servicers who might have otherwise been more aggressive in modifying loans to reduce foreclosures. Surging Foreclosures and Spatial Concentrations of Foreclosed Properties

In many older cities with weak housing markets, but also in some cities with relatively strong economies like Atlanta and Chicago, delinquencies and foreclosures increased well before 2006. By the fitst quarter of 2006, subprime delinquency rates already exceeded 12% in states with more troubled economies, like Pennsylvania, Michigan, Ohio, and Indiana, but also in Georgia and Tennessee, whose economies were still fairly robust at this point.3 However, regions with very hot housing markets experienced low delinquency rates at this point, with California, Arizona, and Nevada having rates below 6%. This was because borrowers struggling with mortgage payments in hot markets could often avoid default or foreclosure by refinancing or selling their homes.

By the summer of 2007, foreclosure rates were accelerating in most large metropolitan areas, with the steepest increases in markets where housing values were declining rapidly, including places like Riverside, CA, Las Vegas, NV, Phoenix, AZ, Sacramento, CA, and Miami, FL (Immergluck, 2008). Surging foreclosures meant that foreclosed properties, which lenders call real estate owned (REO), were beginning to pile up in many metropolitan areas. Slowing housing markets and tightening credit also prevented these markets from absorbing growing numbers of REO properties.

Figure 1 illustrates the growth of estimated total REO properties in several major metropolitan areas between August 2006 and August 2008. I totaled loans in REO status from the Lender Processing Services (LPS) Applied Analytics data set, which combines data from 18 large servicers of mortgage loans, including 9 of 10 ten largest in the nation (LPS, 2009) .4 I then calculated REO densities by dividing these counts for each metropolitan statistical area (MSA) by the number of mortgageable properties (including all buildings containing from one to four dwelling units, plus condominiums) in the same MSA from the 2006 American Community Survey (U.S. Census Bureau, 2008b). To compensate for geographical inconsistencies and incomplete coverage of the total market, I adjusted the REO totals upward based on statewide loan counts from the Mortgage Bankers Association (2009) National Delinquency Survey (NDS; the NDS is the most widely used and cited data source on the national mortgage market).

Figure 1 shows how REOs accumulated in three MSAs whose housing markets had previously been hot (Miami, FL, Riverside, CA, and Las Vegas, NV), as well as in three other MSAs (Cleveland, OH, Detroit, MI, and Atlanta, GA) whose levels of foreclosure and REO activity had been high even before the national foreclosure crisis in 2007. It indicates that the latter had much higher REO densities at the end of 2006, but that REO densities in the formerly hot markets began to grow rapidly as foreclosures surged. REO densities also grew in the cities that began the crisis with already high levels, but not as quickly as in the MSAs experiencing rapid price declines. By late 2007, REO densities in the Riverside and Las Vegas MSAs exceeded those of Atlanta or Cleveland, and by late summer 2008, the Riverside MSA had a higher REO density than the Detroit MSA.

In addition to causing financial and social hardship to individuals and households, high foreclosure rates can have negative effects on neighborhoods and localities, especially when they are geographically concentrated (Apgar & Duda, 2005; Schuetz, Been, & Ellen, 2008). These negative effects, including lower property values, higher crime, and increased costs to municipal government, are expected to be greater if REO properties sit vacant for significant periods of time rather than being promptly absorbed back into the market in some productive way (Mallach, 2009).

Because subprime lending was disproportionately concentrated in minority neighborhoods and because REO absorption may be slow in lower-income neighborhoods if properties are in poor condition and housing demand sluggish, we might expect REOs to be disproportionately concentrated in central cities that are relatively less affluent than their MSAs. However, subprime and high- risk lending also helped fuel fast growth in newer suburban and exurban communities, especially in parts of the Southwest and in California. Media reports suggested that problems may have been disproportionately severe in newly developed communities distant from metropolitan centers (Leinberger, 2008). Research in specific metropolitan areas provides some support for this. Ong and Pfeiffer (2008) examined foreclosures in Los Angeles County in early 2008 and found that exurban location accounted for about 20% of the spatial variation in foreclosure rates. Lehnert and Grover (2008) examined data on subprime loans in the Minneapolis metropolitan area and found that foreclosure rates were relatively high both in some parts of the central city and in some recently developed, exurban communities. Lehnert and Grover's study in particular suggests the possibility that in some places REO accumulation occurred in a doughnut pattern, high in both central city neighborhoods and in outlying suburban or exurban communities. The extent of such a pattern in any particular MSA will likely depend on how much of its suburban or exurban development occurred during the housing boom and the intrametropolitan patterns of home price declines in that MSA.

In order to describe the intrametropolitan REO patterns across large U.S. metropolitan areas, Figure 2 divides more than 8,800 metropolitan zip codes in the 100 largest MSAs into four categories of REO density as of November, 2008. 5 I estimated REO properties in each zip code using the LPS Applied Analytics data, employing the same state-level weights used to generate Figure 1 . I omitted zip codes containing only post office boxes and zip codes with fewer than 500 mortgageable properties. I estimated the number of mortgageable properties in zip codes by adjusting 2000 census counts of such properties using ESRTs 2007 zip code population estimates (ESRI, 2007).

Within each of these REO density categories, Figure 2 also groups zip codes into five intrametropolitan spatial categories. The first spatial category includes zip codes whose land area lies more than 50% within the primary central city in the MSA. The second category includes those zip codes that lie partly, but less than 50%, in the primary central city. The third spatial category includes those zip codes that lie entirely outside the primary central city and are also in the lowest quartile of zip codes arranged by share of residents commuting more than 30 minutes to work by car in 2000. These are labeled, suburb-only, shortcommute zip codes. The fourth category includes those zip codes outside central cities with shares of such commuters in the second or third quartile. The final spatial category includes zip codes outside central cities whose shares of such commuters are in the fourth quartile, and so are called suburb- only, long-commute zip codes.

Figure 2 indicates that zip codes with high- and veryhigh-REO densities are disproportionately located in primary central cities. In the 100 largest MSAs, almost 30% of zip codes with very-high-REO densities and more than 21% of zip codes with high-REO densities are located in central cities, although they represent fewer than 1 8% of all zip codes. The greater REO densities in central city neighborhoods are likely due to the relatively high foreclosure rates in many of these neighborhoods (due in part to racial and spatial concentrations of subprime lending) and possibly to these neighborhoods being slower than other areas to reabsorb REO properties.

By late 2008, some suburban areas also had problems with concentrated REOs. Approximately 68% of zip codes with high densities of REOs and 58% of zip codes with very high densities of REOs were outside central cities. Moreover, suburb-only, long- commute zip codes made up a disproportionate share of the high- and very-high-REO categories, although still a smaller share than central city zip codes. Suburb-only, long-commute zip codes made up over 25% of high-REO zip codes and over 24% of veryhigh-REO zip codes, but only 23% of all zip codes. Meanwhile, suburb-only, low- and moderate-commute zip codes accounted for 49% of zip codes overall, but only 43% of high-REO zip codes and 34% of very-high- REO zip codes. This suggests that, on average, suburban communities located far from job centers have bigger problems with concentrated REO properties than do other suburban areas.

The Federal Response to the National Foreclosure Crisis

Many analysts sounded warnings about worsening foreclosures and weakening housing markets well before 2007. Some, including a number of analysts at the credit rating agencies, warned that falling home prices could spur foreclosures and housing market decline (Immergluck, 2009). Much earlier, consumer advocates and researchers at the U.S. Department of Housing and Urban Development (HUD) documented foreclosure problems associated with subprime lending (Bunce, Gruenstein, Herbert, & Scheessele, 2001). In the spring of 2006, the Consumer Federation of America issued a report warning of the dangers of exotic mortgages (Fishbein &Woodall, 2006). At the end of 2006, the Center for Responsible Lending issued a report forecasting that subprime foreclosures would accumulate to 2.2 million nationwide and that 19% of subprime loans would end in foreclosure (Schloemer, Li, Ernst, & Keest, 2006). Though criticized at the time as alarmist, both predictions proved later to be too conservative. As foreclosures climbed in 2007, federal policy debates over the foreclosure crisis continued. Figure 3 provides a timeline of key events in the central column and defining points in the evolution of the policy response in the righthand column. While there is no universal consensus about precisely when the crisis started, many would point to April 2007, when New Century Financial, one of the largest subprime lenders in the country, filed for bankruptcy. Smaller players in the subprime industry had filed for bankruptcy in preceding months, but the failure of New Century began to reveal the scale of the crisis.