Executive Summary We hold to our view that households are in better shape than widely perceived, nourished by a robust labor market and a formidable supply of pandemic savings. We do not believe that the equity bear market will derail our base-case scenario that consumption will keep the economy afloat over the next several quarters. Empirically, changes in equity wealth have exerted little to no impact on consumption. Housing does have a discernible wealth effect, and consumption may be more sensitive to falling home prices than rising ones. The sharp decline in home prices feared by many investors could prompt homeowners to retrench, realizing the number-one risk to our constructive view. Although home price appreciation is in the process of decelerating, housing remains undersupplied and home prices will not fall precipitously. Housing bubble chatter is unfounded. Consumption Declines Are Few And Far Between
Consumption Declines Are Few And Far Between
Consumption Declines Are Few And Far Between
Bottom Line: Neither the equity bear market nor a softening housing market will stifle consumption. The Fed’s anti-inflation campaign will eventually induce a recession, but wealth effect concerns are overblown. Feature Flush consumers drawing down the mountain of excess savings they accumulated across 2020 and 2021 provide the foundation for our constructive near-term view on risk assets and the economy. Consumer retrenchment is one of the two principal risks to our stance1 and we would abandon it if a meaningful share of households began to cut back. We do not know that households will dip into their savings to keep consuming at something close to their trend pace – the scale of the fiscal transfers that fattened their bank accounts was unprecedented – but we view the low and declining savings rate as providing ongoing validation for our thesis. Households can sustainably dis-save relative to their post-crisis trend (Chart 1), as a 5% savings rate whittles down their remaining $2.1 trillion stash by just $150 billion per quarter. Chart 1An Extended Period Of Dis-saving Is Sustainable
An Extended Period Of Dis-saving Is Sustainable
An Extended Period Of Dis-saving Is Sustainable
The wealth effect is real – household spending fluctuates with wealth – and one may question whether consumers will continue to spend amidst an equity bear market while the 3-percentage-point surge in mortgage rates pressures home values. As counterintuitive as it may seem, however, changes in equity wealth have had a modest and inconsistent effect on consumption. Changes in housing wealth have exerted greater influence, and one study by prominent researchers suggests that the effect is stronger when home prices decline. We consider the empirical evidence regarding equity and housing wealth effects, along with the prospects for a sharp decline in home prices, in this report. What Drives Spending? For all the talk of the wealth effect, consumer spending is predominantly a function of income. Every multi-factor regression we performed (Box 1) indicated that changes in nominal income account for the lion’s share of changes in nominal consumption, with estimates ranging up to 75%. When we regressed real consumption with real income and real measures of equity and housing wealth, the estimates of income’s effect were sharply lower – typically between 10 and 25% – but the modeled results were dramatically less robust. We accordingly focus on the nominal relationships in the rest of this report, though we note that the real regressions reinforced the nominal regressions’ pointed implication that changes in equity wealth are largely irrelevant for explaining changes in consumption. Box 1: A Regression Refresher Multi-factor linear regression is a statistical method for determining which independent variables influence the movements of a dependent variable. Regression analysis reveals the statistical significance of independent variables based on their empirical relationship with the dependent variable. If the relationship is robust enough that it is unlikely to have occurred randomly, the independent variable is deemed to be significant. The regression equation describes a best-fit line that minimizes the individual observations’ aggregate deviation from the line. It includes a constant term, b, marking the point where the best-fit line intersects the y-axis, and an x term that denotes each of the independent variables, paired with a coefficient, a. Each coefficient describes the sensitivity of the dependent variable to changes in the value of each independent variable. For dependent variable y, and independent variables x1, x2, …, xn, the equation is written as: y = a1x1 + a2x2 + … + anxn + b. The robustness of the regression is indicated by its r-squared value, ranging from 0 to 1, which quantifies the share of the dependent variable's movement that is explained by movement in the independent variables. In our research, we used Personal Consumption Expenditures and Personal Income from the National Income Accounts as our measures of consumption and income, respectively. We used the measure of corporate equities held by households and nonprofit organizations from the Fed’s quarterly Financial Accounts of the United States (report Z.1) to measure equity wealth and followed the methodology of Case, Quigley and Shiller (2005 and 2013)2 to calculate housing wealth.3 We also followed Case, Quigley and Shiller’s methodology in regressing the year-over-year percentage change in the natural log of the variables’ values. Homes Trump Stocks Simple regressions, measuring the empirical impact of a single independent variable upon a dependent variable, indicate that changes in equity wealth exert considerably less influence over changes in consumption than changes in housing wealth. With a two-quarter lag, year-over-year consumption has changed by nearly three cents for every dollar move in equity wealth (Chart 2). Three cents are in line with rule-of-thumb estimates, but we note that the regression’s r-squared is less than 3%. An unlagged year-over-year regression posits a 0.6-cent consumption change for every dollar move in equity wealth with a microscopic r-squared of 0.1%. Chart 2Equities' Relationship With Consumption Is Weak And Unreliable, ...
The Wealth Of Households
The Wealth Of Households
The housing wealth regression indicates that every dollar of changes in housing wealth leads to a 38-cent change in consumption. With a 38% r-squared, the housing wealth regression generates a visibly tighter fit (Chart 3), inspiring more confidence in the posited relationship, though it is incomplete without considering any other variables’ role in influencing consumption. The housing wealth relationship is also considerably stronger on an unlagged basis (Table 1). Chart 3... Contrasting With Housing's Stronger, More Consistent Pull
The Wealth Of Households
The Wealth Of Households
Table 1Simple Regression Output
The Wealth Of Households
The Wealth Of Households
Chart 4Equities Are Owned By Low MPC Households
The Wealth Of Households
The Wealth Of Households
It may seem surprising that relatively opaque changes in housing wealth exert a much stronger influence over consumption than immediately observable changes in equity wealth. We think the result is a function of the greater breadth of home ownership; nearly two-thirds of households own their home, and it is far and away the largest asset for all but the wealthiest of families. Stock ownership, on the other hand, is highly concentrated, with the top 1% of households by wealth owning over 50% of equities, and the top 10% owning nearly 90% of them (Chart 4). Fluctuations in the stock market mostly impact households with a low marginal propensity to consume but changes in home prices effect a much fuller sweep of Americans. The simple regressions set the stage for what we discovered when we performed multi-factor regressions, confirming previous researchers’ findings. Income is the primary driver of consumption, with a one-dollar change in nominal income provoking a 65-to-72-cent change in nominal spending, and its statistical significance in the models is beyond question (Table 2). Table 2Multiple Regression Output
The Wealth Of Households
The Wealth Of Households
Equities’ wealth effect is not statistically significant in the unlagged model at a 5% significance level (it’s not even statistically significant at the more forgiving 10% significance level) and it is modest (about 1.5 cents on the dollar) in any event. The model would be better off without including equity wealth as an independent variable. In the model lagging consumption by two quarters, which produces a slightly better fit and accords more easily with our own intuition that wealth effects are not felt instantaneously, consumption moves inversely with equity wealth, falling 3 cents for every one-dollar increase in equity wealth and rising 3 cents for every one-dollar decrease. That result is statistically significant, albeit hard to wrap one’s head around. The housing wealth variable is comfortably significant even at a 1% significance level and its impact is quite large in both the unlagged (14.5 cents on the dollar) and the two-quarter-lagged (11.75 cents on the dollar) specifications. Both model specifications generate high r-squareds, explaining 58% and 60% of the variability in consumption, respectively, and the modeled values fit the actual values extremely well before the pandemic scrambled the relationship between consumption and its drivers (Chart 5). Chart 5A Tight Fit Before The Pandemic
A Tight Fit Before The Pandemic
A Tight Fit Before The Pandemic
We also ran a version of the model that substituted Disposable Income for Personal Income, but it slightly weakened its explanatory power and we judge that the broader Personal Income series is a better input. We also ran a version of the model that used household real estate holdings and mortgage balances from the Fed’s quarterly Z.1 report to calculate a factor that translates gross housing wealth to net housing wealth to reflect that all households do not own their homes free and clear.4 Substituting net housing wealth reduced the model’s explanatory power by about two percentage points but left the individual variables’ significance largely intact while cutting housing’s unlagged and two-quarter lagged wealth effect to 7 and 5 cents, respectively (Table 3). Net housing wealth is more intellectually satisfying than gross housing wealth and the smaller wealth effect estimates are more in line with the peer-reviewed literature. Table 3Multiple Regression Output With Net Housing Wealth
The Wealth Of Households
The Wealth Of Households
Whither Home Prices? Investors appear to be braced for a sizable decline in home prices even though nominal price declines are unusual in the five-decade history of the leading repeat sales price indexes. The Case-Shiller National Index has declined just 19% of the time on a sequential basis and 14% of the time on a year-over-year basis (Chart 6). Excepting the 21 consecutive quarters of year-over-year declines from 1Q07 through 1Q12, the Case-Shiller National Index has declined in just five quarters over 41 years, all during the 1990-91 recession that featured tax law changes sharply curtailing individuals’ ability to benefit from losses on real estate investments. The FHFA (née OFHEO) House Price Index has declined on a year-over-year basis just 11% of the time, with only one decline occurring outside of 2007 to 2012 (Chart 7). Chart 6Ex-The Crisis, Declines Are Rare, ...
Ex-The Crisis, Declines Are Rare, ...
Ex-The Crisis, Declines Are Rare, ...
Chart 7... In Both Major Series
... In Both Major Series
... In Both Major Series
Investors expecting a decline therefore appear to be anchoring to an extreme outlier. We cringe whenever we hear the term “housing bubble” used to liken today’s backdrop to the one that preceded the financial crisis. Make no mistake: it is not 2007 in the housing finance market in any way, shape or form. Residential mortgage originations have been made to vastly better borrowers than they were in the run-up to the crisis (Chart 8) and they’ve been made on far more solid terms, as the loan-to-value ratio for residential mortgages has shrunk by 25 percentage points in the immediate aftermath of the bust to its easily sustainable levels of the early ‘80s (Chart 9). Chart 8Mortgages Have Been Extended To Better Borrowers ...
The Wealth Of Households
The Wealth Of Households
Chart 9... On Better Terms Than Before The Crisis
... On Better Terms Than Before The Crisis
... On Better Terms Than Before The Crisis
Chart 10Housing Supply Is Tight
Housing Supply Is Tight
Housing Supply Is Tight
Housing is broadly undersupplied, as evidenced by the record-low home vacancy rate (Chart 10). Higher mortgage rates have surely put monthly payments out of the reach of some aspiring buyers, sending them to the sidelines, but supply remains constrained and home prices fall slowly. Kahneman and Tversky demonstrated that people are quick to take gains by selling appreciated assets but slow to part with assets that are under water. Even if we are underestimating the eventual magnitude of a decline in home prices, we are confident that the decline will not be sudden. Homeowners with discretion over when they sell will wait to exercise it; turnover will slow as pricing softens and the reduced supply will help to mitigate the declines. Investment Implications We were inspired to explore the housing wealth effect by a striking assertion featured in a leading market periodical two weeks ago. An independent strategist stated that the wealth effect from a one dollar decline in home prices was a whopping 40 cents, while the effect of a like decline in equity prices was 10 cents. The assertion was passed on without comment or criticism by the publication, which has long touted its skepticism and unwillingness to accept bullish statements at face value. Alas for its readers, the standard apparently does not apply to bearish claims, no matter how far off the beam they may be. (Based on our results, we suspect these wealth effect estimates are based on simple regressions.) Divergent views are what make a market, but nothing in the body of peer-reviewed research supports the idea that the $6.5 trillion decline in directly owned equities and a hypothetical 10% decline in home equity from its nearly $30 trillion June 30th level will extinguish $650 billion and $1.2 trillion of consumption, respectively. That nearly $2 trillion hit would be punishing, given consumption's current $17 trillion annualized pace. It would also be unprecedented: since the Personal Consumption Expenditures series began in 1950, nominal consumption has only ever declined by a margin that can be seen by the naked eye during the Great Recession and the COVID pandemic (Chart 11). Those historic declines amounted to 3.5% from the 3Q08 peak to the 2Q09 trough and 11.4% from the 4Q20 peak to the lockdown 2Q21 trough. Chart 11Visible Declines In Nominal Spending Are Rare
Visible Declines In Nominal Spending Are Rare
Visible Declines In Nominal Spending Are Rare
We are only too happy to take the other side of the view that another 11% decline could be in store, assuming the absence of nuclear war or another pandemic. We think the 3.5% Great Recession decline will likely remain out of reach, as well, given that the financial crisis emerged from a concatenation of events that cannot repeat now that regulators have so thoroughly clipped the banking system’s wings. Not every investor subscribes to Chicken Little warnings about the housing market, but the promiscuity with which the term bubble is thrown around strongly suggests to us that the consensus view overestimates the probability of a dire economic outcome. When subsequent events reveal that the shock probability has been overstated, the consensus economic and S&P 500 earnings views will have to be revised upward and we believe the eventual revisions will provide risk assets with a path to recover some of the ground they’ve lost this year. We continue to believe that it would be premature to implement full-on defensive asset allocation measures before they do. Doug Peta, CFA Chief US Investment Strategist dougp@bcaresearch.com Footnotes 1 A breakout in long-run inflation expectations is the other. 2 Case, Karl E., John M. Quigley, and Robert J. Shiller, “Comparing Wealth Effects: the Stock Market versus the Housing Market,” Advances in Microeconomics, 5(1),2005: 1-32. Case, Karl E., John M. Quigley, and Robert J. Shiller, “Wealth Effects Revisited: 1975-2012,” NBER Working Paper 18667, January 2013. 3 Case, Quigley and Shiller calculate housing wealth in time t, HWt, as the product of the number of US households, Nt, the homeownership rate, ORt, the average price of a single-family home in the base period (1Q75 in our study), AVGBASE, and a weighted repeat sales price index relative to its base period value, (PIt/PIBASE). We used the National Association of Realtors’ average existing home price series and the Case-Shiller National Index for variables AVG and PI, respectively, as per the following equation: HWt = Nt × ORt × AVG1Q75 × (PIt/PI1Q75) 4 HWt, described in the second footnote, is a gross measure of housing wealth. We divided outstanding mortgage debt by the value of households’ real estate holdings to calculate the aggregate residential mortgage loan-to-value ratio, LTV. We subtracted LTV from 1 to calculate the share of housing value that represented households’ aggregate home equity and multiplied it by HWt to produce an estimate of net housing wealth, NHW: NHWt = HWt × (1 – LTVt)