Although U.S. home prices are enjoying stable economic support, the current record pace of price growth is not sustainable for an extended period. While the most recent consumer price inflation (CPI) readings have been the highest in years, S&P Global Ratings believes that, under the Fed's currently announced policy settings, inflation should return to the target rate of 2% on average, likely by year-end 2022 (although we expect the level of consumer prices to be permanently higher).

Regardless of how long it persists, however, the CPI is substantially lower than the pace at which home prices are climbing. This suggests that housing will only become more unaffordable in the near term. It is therefore reasonable to ask whether monthly year-over-year Case-Shiller National Home Price Index (HPI) changes (i.e., house price appreciation, or HPA) will settle into a rate consistent with the historical norm of roughly 3%–5% annual growth (i.e., a soft landing), or whether there will be a correction that could lead to outright price declines at the national level (i.e., a hard landing). To address this question, we: (1) examine the drivers of HPA in the current market and contrast them with some of the features of the pre-Global Financial Crisis (GFC) market; (2) develop a regression to provide downside scenario analyses; and (3) assess the impact that sudden home price declines might have on certain U.S. non-agency residential mortgage-backed securities (RMBS).

Home Prices Are Increasing At A Record Pace

In the years leading up to the GFC, the Case-Shiller National Home Price Index (HPI) saw record high year-over-year increases, with the greatest change in September 2005 at +14.4%. During 2021, the monthly year-over-year HPI increases (i.e., HPA) have also been in the double digits, exceeding the September 2005 figure in April at +15.0%, and subsequently hitting new records of +16.9% in May, +18.7% in June, +19.8% in July, and +19.9% in August. Although still high, the September figure was the first reading to break the upward trend, coming in at +19.5%, suggesting that perhaps a softening in HPA is imminent. Chart 1 shows the historical monthly paths of both the home price index and its annual appreciation.

Chart 1

Comparing The Pre-GFC And Current U.S. Housing Markets

In our view, many features characteristic of the pre-GFC housing market are absent from the current market. Most of the funding for residential mortgages is now facilitated by the government-sponsored enterprises (or simply "agencies") Fannie Mae and Freddie Mac, which tend to be uniform in their underwriting guidelines. Meanwhile, regulations such as the Ability-to-Repay (ATR) and risk retention rules, which aim to encourage responsible lending, are standard features of the (now much smaller) non-agency market. Mortgage products have changed, too. The share of adjustable-rate mortgages (ARMs) originated in 2020 is lower than it was pre-GFC due primarily to the currently low mortgage rate environment, but also attributable to the disappearance of teaser-rate features on mortgages to subprime borrowers (see "How U.S. Structured Finance Has Changed Since The Credit Crisis," published Feb. 11, 2020). Chart 2 shows how several economic indicators and mortgage-related characteristics measured in 2020 differ from their values as reported in 2005, when the housing market was near the peak of its pre-GFC price trajectory.

The Drivers Of Record HPA

We believe the following are the main drivers of the HPA in the current market:

• Demand from millennials;
• The low interest rate environment;
• The lack of housing supply;
• Building costs;
• COVID-19 pandemic effects; and
• The strength of the consumer.

We examine these drivers of HPA in the current market in detail below.

Demand from millennials

As the demographic group aged 25-44 years is most likely to buy houses, an increase in the population of this group is expected to be correlated with the demand for housing. The growth rate of this age bracket peaked in the late 1980s when the baby boomers (roughly representing those born between 1946 and 1964) started to exit the homebuying target age range. Population growth of this segment did not pick up again until after 2010, when millennials (roughly representing those born between 1982 and 2000) started to enter their homebuying years.

Millennials are now the largest generation, having recently overtaken the baby boomers, and since 2014, they have been the largest share of homebuyers in the country. After years of delayed entry into the housing market, due in part to financial constraints, such as student loan debt and limited job opportunities during and after the GFC, millennials are now forming families and making home purchases. This is a major contributor to the strong demand for housing that is currently driving up home prices.

The low interest rate environment

The low interest rate environment is also driving HPA. According to the Freddie Mac Primary Market Survey, the 30-year fixed-rate mortgage (adjusted for points and fees) hit a record low of 2.65% in January 2021. This benchmark rate remains near 3%, enabling homeowners to borrow at relatively low costs and keep monthly payment obligations at more affordable levels than they would otherwise be in a higher interest rate environment. There may also be a type of "media effect" at play as one is consistently reminded of record low mortgage rates. Indeed, the "fear of missing out" (FOMO) (https://www.npr.org/2021/05/21/999093282/record-prices-for-red-hot-housing-market-fear-of-missing-out) phenomenon could be incentivizing prospective buyers to make a purchase before an ostensibly unique opportunity is lost.

The Fed will likely maintain its target range for overnight borrowing rates near zero until mid-2022 and raise rates gradually thereafter. Longer-term bond yields are expected to remain low as the Fed is unlikely to shrink its balance sheet (by rolling-off maturing assets or outright sale of assets) for some time after tapering of its scheduled quantitative easing is completed in June (or even earlier as risks of higher inflation nudge the Fed to accelerate tapering). The 10-year Treasury note yield, which is correlated with the 30-year conforming mortgage rate, sits below 1.5%. This is up from the July 2020 record low, but still firmly below the long-term average. While the Fed has signaled that the time to commence tapering its asset purchases is drawing near and that it would end by mid-year 2022, complete balance sheet normalization is still years away (see "Complete Fed Balance Sheet Normalization Is Still Years Away," published Aug. 16, 2021). We believe the Fed is likely to follow its quantitative easing exit playbook from the last monetary cycle. That is, it will keep the size of its asset holdings roughly constant (by reinvesting maturing assets) for some period after tapering, and then begin a period of passive asset roll-offs as the bonds mature. These factors suggest that the low mortgage rate regime will be with us during the near term. S&P Global Ratings forecasts that the 30-year fixed-rate mortgage will increase to 3.4% in 2022 and then to 3.7% in 2023 (see "Economic Outlook U.S. Q1 2022: Cruising At A Lower Altitude," published Nov. 29, 2021).

Housing supply is lagging

While the currently strong pace of HPA is a recent phenomenon, home prices have been appreciating since 2012. Since prices began rising from GFC-era lows, lack of adequate home supply has consistently been one of the primary factors putting upward pressure on home prices. Part of the problem is that there was an oversupply of houses during the years leading up to the GFC. The U.S. homeownership rate hit a record high of near 70% in 2004, and then fell steadily until bottoming out in 2016 (chart 2 contrasts the rate in 2005 with that in 2020). At its peak in 2009, there was a home supply (measured as the ratio of total houses for sale to the number of houses sold per month) of over 11 months, according to data from the U.S. Census and the U.S. Department of Housing and Urban Development (HUD). Because some of this supply had to be absorbed in the post-GFC years, there was a relative reduction in the number of houses being built, and supply returned to a level of five to seven months, which is likely insufficient to satisfy the demand from the current market.

After new privately owned housing starts fell to a record low in 2009, they have since increased steadily. S&P Global Ratings estimates that there will be 1.6 million starts in 2021, and that this figure levels out at 1.5 million in 2022 and 2023, despite the demand from the millennial segment of the population (see "Economic Outlook U.S. Q1 2022: Cruising At A Lower Altitude," published Nov. 29, 2021). According to the National Association of Realtors, millennials continue to make up the largest share of homebuyers at 37%, broken out as younger millennials (aged 22-30 years) at 14% and older millennials (aged 31-40 years) at 23%.

Although demand for housing has been increasing with the formation of millennial households, the past decade marked a period of underbuilding in the residential housing market relative to the historical average. Chart 3 shows: (1) the total completions of new residential structures over the past 10 years, (2) the 10-year growth in the total number of households, and (3) the ratio between the two. This ratio, which peaked in 2010 and has since receded to the lowest level in over 40 years, suggests that the supply of new homes is not yet keeping up with the growth of households. Another important measure of demand is the homeowner vacancy rate, which currently sits below 1%, well below the record high of 2.9% in 2008, suggesting that demand is outweighing supply.

Chart 3

As first-time millennial homebuyers find themselves priced out of expensive markets, they are turning to more affordable ones—not coincidentally, the same markets to which many companies are relocating to or establishing themselves in. Homes in cities such as Boise and Austin have recently experienced substantial upward price pressure. This in turn has caused residential real estate markets in these regions to become overvalued according to S&P Global Ratings' over-/under-valuation analysis, which examines local affordability relative to the long-term average. This is not to say that they are the most expensive markets in the U.S. Rather, these markets are now less affordable than they have been over the past 15 years. We estimate that the broader U.S. market is currently overvalued by 5% (see "Shift From Under To Overvalued: What It Means For U.S. RMBS," published Oct. 29, 2021). In addition to demand from corporations and millennials seeking starter homes in secondary markets, these relatively overvalued regions have been attracting individuals that were priced out of larger markets, such as California and the Pacific Northwest.

Building costs

Adding to the supply-demand imbalance is the increased cost of building new homes. Contributing to the problem is the continually increasing cost and scarcity of labor. The labor shortage has been an issue for more than a decade, starting when demand for new homes dried up after the GFC and many laborers left the homebuilding industry. Unsurprisingly, the decades-long push for young people to go to college has driven down trade-school enrollment, which in turn has deprived builders of skilled labor. Also, the declining numbers of construction workers in the U.S. has rendered builders short of unskilled labor. The result of this trend has been an extension of construction times, which we believe has slowed the pace of new homes coming to market. In some cases, builders have had to pay more for laborers, and the additional costs are presumably being passed on to the homebuyers.

The increased cost of building new homes is also attributable to the rising cost of construction materials, such as lumber. At the onset of the COVID-19 pandemic in late March 2020, the price of lumber futures was roughly $300 per thousand board-feet. The price of lumber futures increased to a record high of close to $1,700 in May 2021, but has since fallen substantially. Nevertheless, supply chain disruptions could cause prices of finished housing products (as opposed to raw materials) to remain high and contribute to both the sustained high cost of building new homes and the delay of completions in the near term.

Homes sold today are often sitting on land purchased two to three years prior. According to John Burns Consulting, there was a 15% year-over-year increase in lot prices in first-quarter 2021. In the second quarter, lot prices were up over 16% year over year. This means that as homebuilders buy land today for use in the future, they will likely need to raise prices to maintain margins. By region, the Midwest saw the smallest increase in lot prices in the second quarter at about 9.5%, relative to an approximate 25% increase in lot prices in the Southwest.

Finally, there are physical limitations to available land for development. Areas bordered by mountains or bodies of water (the island of Manhattan is an example) present physical impediments that will continue to limit construction and similarly constrain supply as the population grows. The U.S. population grew roughly 10% between 2005 and 2020 and now sits at just over 330 million. While the pace of U.S. population growth is slowing, the United Nations expects it to increase to 340 million by 2025 and to 350 million by 2030. Also, the homebuying segment aged 25–44 years increased to 89 million in 2020 from 83 million in 2005, as shown in chart 2. This segment is expected to increase to 92 million in 2025 and to 94 million in 2030 (forecasts from the United Nations). Moreover, as people enjoy longer lives, they also stay in their houses longer, which further limits the supply of existing homes.

COVID-19 pandemic effects

Global lockdown measures, put in place to control the spread of the pandemic, encouraged a subset of the population to move from urban areas to suburban areas as households sought additional space and "social distance." This "de-urbanization" is putting upward pressure on housing demand, largely from the subset of millennials who have typically been renters and are now switching to homeownership as they form families. In a post-pandemic society, it is expected that more people will work remotely, at least on a part-time basis (although the portion of the work force that will do so remains unclear). This suggests that many people will desire larger living spaces and will need appropriate floorplans to facilitate a proper balance of work- and home-life. Nevertheless, the transition to buyer from renter can be prohibitively expensive. Forty years ago, 40% of newly constructed homes could be described as entry-level. As of 2019, the share sits near 7% (https://www.npr.org/2021/09/04/1033585422/the-housing-shortage-is-significant-its-acute-for-small-entry-level-homes).

Consistent with the supply-demand imbalance across the U.S., this scarcity of affordable dwellings for first-time buyers could continue to drive prices higher.

The strength of the consumer

In addition to improved controls in mortgage lending and a shift away from a supply-driven market to a demand-driven one, the U.S. consumer is in a relatively strong position today compared to the pre-GFC period. According to the Federal Reserve's Z1 report, U.S. households ended 2020 with $129 trillion in total assets, up from $122 trillion in 2019 and sitting at $151 trillion as of second quarter 2021. Part of the growth is attributable to the current interest rate environment and fiscal policies put in place to address the pandemic. As of second-quarter 2021, household financial obligations as a percentage of disposable personal income were 13.8%. This is substantially lower that the series high of 18.0% during fourth-quarter 2007, on the eve of the GFC. (For details on this measure of consumer health, see, "https://www.federalreserve.gov/releases/housedebt/about.htm".)

Meanwhile, existing homeowners are growing their home equity at an unprecedented rate. A report from CoreLogic with data through second-quarter 2021 indicates U.S. homeowners with mortgages have seen their equity grow by $2.9 trillion since second-quarter 2020, marking a year-over-year increase of 29.3% (https://www.corelogic.com/intelligence/homeowner-equity-insights/). (For more on the health of the U.S. consumer, see "U.S. Structured Finance Snapshot: The Health Of U.S. Consumers," published Sept. 13, 2021.) Finally, borrowers generally have larger equity positions in their homes relative to the pre-GFC period.

Regressing The Dependence Of HPA On Economic Factors

The framework

To understand how home prices depend on observable economic factors, we estimated the linear relationships (assuming they exist) between changes in the S&P Case Shiller National Home Price Index (i.e., HPA) and various independent covariates. We carried out a regression using quarterly data from 1988 through fourth-quarter 2019, which was immediately prior to the onset of the pandemic in late first-quarter 2020.

To account for time-series relationships within the data and the potential for serially correlated error terms, we used a linear regression with ARIMA (2, 1, 3) errors. Because first differences are applied to HPA as well as the covariates, the data were rendered stationary. (For more information on this technique, which was carried out with software package R, see Forecasting: Methods and Applications by Makridakis, Wheelwright, and Hyndman, 3rd ed., John Wiley and Sons, 2005.)

Once calibrated, the regression allows one to understand the magnitude and direction of relationships between HPA and each of the standardized covariates. If a particular covariate is estimated to have a positive parameter, HPA tends to move in the same direction as that covariate, and vice versa. The final regression framework was chosen in part because it led to certain optimal fit statistics, but more importantly, because we believe it captured the essential drivers of HPA. Once calibrated to historical data, the regression generates forecasts based on assumed paths of the covariates.

While such a regression captures the historical dependence of home price changes upon related market and economic factors, it cannot explain behavior that is not directly observable, such as ostensibly irrational mass consumer behavior. Moreover, if an economic environment shifts to a new, previously unobserved state (i.e., a regime shift), the historical relationships may not persist.

The covariates, which included either the factors discussed above or other related measures, were in some cases lagged by up to three quarters. We regressed the quarter-over-quarter change in HPI (i.e., quarterly HPA) against the following covariates and determined that all parameters had signs consistent with the expected impact of the associated covariates.

• The 30-year fixed mortgage rate: This economic variable appears twice, lagged at one and two quarters, respectively. The parameter estimates were negative.
• New home sales: This measure of housing demand is not lagged. The parameter estimate was positive.
• Household assets: This measure of consumer wealth is lagged by one quarter. The parameter estimate was positive.
• Private residential fixed investment (PRFI): This demand term comprises purchases of private residential structures and residential equipment, which is owned by landlords and then rented to tenants. It was lagged by three quarters. The parameter estimate was positive.
• Housing completions: This measure of supply was lagged by one quarter. The parameter estimate was negative.
• Share of population aged 25–44 years: This variable was not lagged. Quarterly data were obtained via cubic spline interpolation of annual data. The parameter estimate was positive.

With the exception of population data, which are from the U.N., economic data were downloaded from the Federal Reserve Bank of St. Louis economic research (FRED) website (https://fred.stlouisfed.org/).

Regime shift

The current economic environment is in a state of flux as the pandemic evolves and society adapts to various market changes in the U.S. This means that the pandemic may have induced a temporary economic regime shift that no economic variables could adequately predict. Indeed, some of the factors driving home prices are difficult to quantify and are not easily captured in standard regression framework. While we expect supply chains to return to a state of normal functioning and the broader labor market should in time revert to its former strength, some of the factors affecting housing demand, such as the need for more living space, may persist for an extended period. To avoid calibrating the regression with data that might distort long-term relationships established from the pre-pandemic period, we withheld seven quarters of data (first-quarter 2020 through third-quarter 2021). Under the assumption that the regime will revert to its pre-pandemic structure in the future, the longer-term predictive power of the model could be improved by withholding these data.

Covariate paths

With the regression calibrated to historical data, we carried out two scenario analyses in which future paths for each covariate were assumed. Because we are interested in potential downside cases, we contemplated adverse paths that were not necessarily consistent with any current economic forecasts, but instead coincided with movements in covariates that served to slow down HPA. That is, a covariate with an estimated positive parameter sign would be assigned a monotonically decreasing path, and vice versa.

In both scenarios, paths were chosen such that either: (1) a global covariate minimum/maximum was achieved, or (2) there was an increase/decrease between fourth-quarter 2021 and third-quarter 2026 that was consistent with historical behavior over comparable periods. In the more benign scenario (scenario 1), a linear path was used, whereas in the more extreme scenario (scenario 2), a nonlinear but smooth path was used. In both scenario paths, the terminal covariate value was the same. The appendix shows the hypothetical covariate paths.

Forecasts

The HPA forecasts based upon these scenarios are presented in charts 4 and 5. In both, we show the regressions overlaid with historical data through fourth quarter 2019. From first quarter 2020 through third-quarter 2021, we show how the actual (withheld) seven quarters compare to estimated HPA. While the regression results undershoot somewhat, they are close and generally within the 80% prediction intervals (not shown here). We also present the historical and forecast paths of HPI (as determined from the quarterly HPA). In both cases, prices eventually start to fall, but not until over a year has passed. In the first scenario, prices start to decline after fourth-quarter 2024 and hit a 9% year-over-year price decline by third-quarter 2026. In the second scenario, prices start to decline after first quarter 2023 and hit 14% year-over-year price decline in first-quarter 2026.

Chart 4

Chart 5

Interestingly, even under the more extreme stress that we considered, HPA hits (3.7)% quarter-over-quarter only in 2026, which is shy of the (3.8)% reading in first-quarter 2009 during the GFC. This is not necessarily surprising given the current trajectory of home prices. Moreover, aggregate home price measurements tend to enjoy a certain degree of momentum, which our regression formulation captures. As such, instantaneous price movements would generally not be expected.

The regression considers only economic and housing market predictors. There is no direct attempt to capture population behavioral characteristics that could be responsible for the current trend in HPA. Because it is difficult to quantify the extent to which such qualitative attributes may be responsible for the current trend in HPA, phenomena such as FOMO have no direct influence on our forecast HPA paths. This is perhaps the greatest potential source of error in the analysis and is why we withheld the seven quarters starting with the onset of the pandemic, when these forces may have been more pronounced.

The Impact On U.S. RMBS

We examined the potential ratings impact on U.S. non-agency RMBS should home prices fall suddenly. We considered three different scenarios in which prices fall by either 5%, 10%, or 20% at the national level. Our analysis assumes an immediate drop shortly after the consummation of two hypothetical post-2008 RMBS transactions: one transaction prime jumbo, the other non-qualified mortgage (non-QM) (for details on various RMBS products, see "How Do Non-QM Loans Stack Up Against Pre-Recession Mortgage Products?" published Dec. 4, 2017). The potential impact on ratings derives from both an increase in borrower default expectation (given a reduction in home equity) and reduced recovery values (given a lower disposition price of the property) as applied when evaluating certain RMBS using our credit model LEVELS.

Table 1 lists the capital structure of proxy prime and non-QM transactions with ratings (prior to hypothetical price declines) and the new potential ratings that could result from assumed instantaneous price declines.

Table 1

One should consider that home price declines would likely not be immediate as we have assumed for purposes of our hypothetical scenario analysis. Rather, a downturn in HPI would likely transpire over the course of months, if not years, as illustrated in the regression discussed in the prior section. Moreover, such an immediate drop in home prices would be accompanied by property value declines associated with various rating scenarios (e.g., for a 'BBB' level of stress it is assumed that the liquidation value of the property is subject to a further reduction of about 23%, as specified in our criteria, beyond the 5%, 10%, and 20% home price decline scenarios). To put things in perspective, the steepest decline we used in the RMBS impact scenario analysis (a 20% instantaneous price drop) was substantially more severe than the 13% annual decline observed during the GFC (see chart 1) and would be expected to take more than a year to transpire. The potential rating transitions under such scenarios would be within our credit stability deterioration thresholds (provided in "S&P Global Ratings Definitions," published Nov. 10, 2021) as the time to reach the 20% decline would be expected to exceed one year.

The scenarios show similar rating movements when comparing prime and non-QM RMBS, although under the larger home price decline scenario, non-QM RMBS may have less movement at the middle of the capital structure as compared to prime. This is likely due to the soft credit enhancement provided by excess interest (typical for most non-QM transactions), which is less correlated to changes in the loss profile of a pool that would come about from home price declines and higher borrower leverage.

Our analysis focused only on the impact of home price declines and not on the behavioral characteristics of the pools under different conditions (e.g., delinquency rate changes). Although we assumed credit enhancement levels to be unchanged relative to the close of the securitization and the subsequent instantaneous decline in home prices, some degree of prepayment activity and deleveraging of the capital structure would be expected to increase enhancement levels in the near term, particularly for more senior bonds. We further note that home price declines may lead to a reduction in the overvaluation of both regional markets and the U.S. broadly. This in turn could result in lower rating-level specific loss severities than we considered in our scenario analysis.

A Soft Landing For HPA

The rapid increase in home prices across the U.S. is attributable to a combination of causes, many of them economic and connected to supply/demand fundamentals. Indeed, the housing market has experienced a confluence of positive factors, including historically low mortgage rates, strong demand from a growing population of individuals entering their home-buying years, and insufficient inventory to meet this demand. More difficult to quantify, but perhaps equally important in driving up home prices, have been the indirect effects of the pandemic. The desire to own residential real estate has generally intensified as renters and owners of small dwellings sought additional living space to accommodate pandemic and post-pandemic work/school arrangements and, in some cases, second homes outside densely populated cities. We remind the reader that the framework adopted here does not capture the dependency of home prices on behavioral factors. Also, it assumes a reversion to a pre-pandemic state, which may not come about in the near term or even later.

There are similarities between the current pattern of HPA and the 2004–2006 experience. However, differences in the fundamental drivers of HPA then and now suggest the current environment is a healthier one than the housing bubble that preceded the GFC. The third quarter, seasonally-adjusted, Case-Shiller HPA reading of +4.75% quarter-over-quarter was down slightly from that of the prior quarterly reading (when it was +5.17%). While this suggests that the recent upward trend in HPA might be attenuating, the question remains: when will HPA revert to a normal pattern, and how volatile will this transition be? The regression analysis and associated downside forecasts suggest that relatively adverse movements in predictors of HPA are unlikely to cause prices to fall in the near term. Nevertheless, if prices were to fall dramatically, our analysis suggests that the impact on certain non-agency RMBS transactions would likely be limited to two rating category movements (e.g., 'AAA' to 'A').

Appendix: Covariate Paths

For five of the six covariates, we have displayed the hypothetical trajectories (from scenarios 1 and 2) on the graphs in charts 6–10. In the case of the population share aged between 25 and 44 years, we assumed the same hypothetical path for both scenarios because these are unlikely to change course, regardless of the economic environment (see chart 11). Using the U.N. forecasts for the years 2025 and 2030, both for the homebuying population and the U.S. population, we interpolated using cubic splines to generate quarterly estimates.

Chart 6

Chart 7

Chart 8

Chart 9

Chart 10

Chart 11