To make sense of elevated interest rates, rolling recessions, supply issues and a tech revolution, you may need to apply a new investing framework.
Whither the U.S. economy?
So far in 2023, it hasn’t entered a recession and what’s more, the old “recession template” for investing—selling stocks to buy bonds and hold more cash—hasn’t been especially helpful. It’s easy to see why: GDP growth (not adjusted for inflation) has stabilized at 7.0%–7.5%. Meanwhile, operating profit margins for S&P 500 companies have stabilized, too, on average, modestly above their pre-pandemic levels (chart below).
These are not recessionary dynamics—they are more consistent with an economy running a bit too hot.1 No wonder risk assets have performed well year to date (YTD). The S&P 500 Index is up by more than 15% and high yield bond spreads have narrowed as of mid-July 2023.2 Here is what we see emerging: Beneath the surface of broad market metrics, some industry sectors are weakening while other sectors are experiencing strong demand amid shortages. Equity strategists have a term for this phenomenon: “rolling sector recessions.” Typically, these occur beneath the surface of broad markets, which is consistent with the patterns we identify and assess below.
Navigating industry-specific rolling recessions as an investor in a macro economy that keeps growing robustly in nominal terms (excluding inflation) requires a careful and nuanced analysis.
Our aim is to provide a framework for understanding these stark sectoral differences and—in the final section—explore how artificial intelligence (AI) fits in, as it begins to impact the real economy and drive corporate earnings. No one yet knows exactly how this technological revolution will play out, but the trend lines are becoming clear: AI-related stocks are up 37% YTD as of July 2023.4
Investing in a world of shortages and excesses
What market forces are at work in the sectors being hit by shortages and oversupply—and more importantly, how can a concept of comparative shortages and excesses help inform portfolio strategy? Let’s examine the fundamental drivers affecting U.S. housing, autos, regional banks and commercial real estate (specifically, offices).
Their differences are more obvious than their similarities, but housing and autos are linked because these two sectors are still operating in a state of shortage: Their current inventory levels are insufficient to meet strong demand. Conversely, regional banks and office buildings can be characterized as struggling with an excess of supply.
Supply shortage: Housing
We recently estimated the United States has a shortage of 2 million to 2.5 million housing units relative to population pressures. We reached this number by examining housing starts and completions relative to trend household formation growth (and also accounting for obsolescence).5
This shortage has sharply lifted the value of new-build housing supply, especially with high mortgage rates deterring existing homeowners from selling. The current market rate for a 30-year fixed mortgage is just under 7%, which is significantly higher than the average mortgage rate of 3.5% that U.S. homeowners currently have locked in across the economy. For these reasons, new home sales continue to make up a higher fraction of sales than in the 2010s.
None of these market factors suggest that we will see a comparable 1950s-style housing construction boom, but the general direction remains the same. The only way to address the current housing shortage—and reduce the age of the U.S. housing stock—is to permit more construction. We see evidence this is underway, making housing a likely secular outperformer in the 2020s, and putting an end to the sector’s decade of meaningful underperformance in the 2010s.
Housing sector equities (i.e., homebuilders) do not look particularly expensive. Despite the sector’s strong performance YTD, housing stocks trade at a 1.6x forward price/book ratio. That ratio is a 16.5% discount vs. the average differential vs. the S&P 500 Index over the past 10 years.
Supply shortage: Autos
The auto shortage began in the summer months of 2020 as a result of COVID-19 lockdown measures and ensuing supply chain disruptions. Then, in 2021, a worldwide shortage of semiconductors further curtailed new automobile production. As soon as the U.S. government lifted travel restrictions, consumers took to the road and the used car market, which was the main source of supply, boomed. (Government fiscal and monetary stimulus, as well as a surge in regional migration patterns fueled those car purchases.)
Quantifying the auto shortage is trickier than quantifying the housing shortage, but we can usefully compare auto sales prior to the pandemic vs. what transpired throughout the pandemic. Pre-pandemic, about 17 million U.S. light vehicles were sold annually; since then, 14.5 million were sold in 2020, 14.9 million in 2021 and 13.8 million in 2022.
So, assuming the pandemic never hit—and growing the run-rate forward—the data would imply a cumulative under-shipment of almost 8 million light vehicles. To be sure, this is an simplification that likely overstates the shortage because it doesn’t account for higher sustained auto input costs, which have affected sales in recent years.6
Another way to assess the auto shortage involves inventories sitting on dealer lots. The chart below shows inventory as the number of selling days that it would take to exhaust dealer supply. Prior to the pandemic, dealers normally held about 60–70 days’ worth of supply on their lots—the current level is 20. In our view, we expect the industry to reach a new equilibrium of approximately 45–55 days’ worth of supply by mid-2024. Expected improvements in supply chain productivity (such as reduced vehicle complexity and greater automation in the production process) mean that the sector is unlikely to return to its pre-pandemic equilibrium.7
When it comes to the valuation of auto equities, Tesla, which currently trades at a 65x forward price/earnings (P/E) multiple, stands apart. Tesla’s high multiple can be attributed to the company’s growth potential in the context of the global energy transition from fossil fuels to renewables and the concomitant shift to electric vehicles.8 Ford Motor Company and General Motors (whose stocks are up 33.8% and 13.5% YTD through mid-July, respectively) are not trading at expensive valuations—much like what we see in U.S. housing. Ford trades at a forward P/E discount of 3.5%, whereas General Motors’ P/E discount is 22.5%.9
As for where the auto and housing sectors may be headed, it’s also useful to note how quickly pandemic-related supply shortages were alleviated in household products (i.e., goods produced by the likes of Proctor & Gamble, Clorox and Colgate-Palmolive) vs. housing and autos.
Gross profit margins for these sectors reveal the dynamics in play: Margins rose in each as a result of the global pandemic and associated shortages, but—because of the greater difficulty in alleviating supply shortages in housing and autos—margins are still well above their pre-pandemic levels (chart below). In household products, by contrast, ramped up production (combined with cooling demand) alleviated shortages quite rapidly, resulting in margin compression to a level below pre-pandemic norms.
Looking ahead, we expect to see auto margins normalize more quickly than housing margins. Notably, we wouldn’t be surprised to see housing margins remain above their pre-pandemic levels well into 2024 and possibly even into 2025.
Excess supply: Office buildings
That mismatch continues to put upward pressure on vacancy rates. Although we were likely to see some excess supply in this sector even if the pandemic had never hit (given the impact of property obsolescence), the onset of the global health crisis, the abrupt shift to remote work and layoffs in the technology industry have all weighed heavily over the past year. All are now contributing to the collapse in market values for REITs that have high exposure to offices.
Not all office assets are created equally, however: Aging properties in downtown central business districts (CBDs) in major U.S. cities are far more affected than suburban office assets. While office CBD vacancy rates continue to rise, suburban office vacancy rates have actually been falling in recent quarters (chart below). Offices make up approximately 15% of all commercial real estate, and CBD makes up about 60% of office, so about 9% of all CRE is challenged by escalating vacancy rates as a result of the pandemic shock. This dynamic underscores the sector-specific—or micro—nature of the problem.11
Excess supply: Regional banks
The U.S. banking sector is also experiencing something akin to excess supply. Despite a dramatic decline in the number of U.S. banks over the past 30 years—the total has dropped by more than 70% since the early 1980s—the U.S. remains an outlier in a broader, international context. Our analysis in the chart below, which takes into account the number of banks in the system as well as deposit concentration, illustrates just how much of an anomaly the U.S. is, internationally.12
What is driving the wave of U.S. bank consolidation that began this spring? Uninsured deposits in the system are raising operational risks; changes in technology are making it easier than ever to move funds instantly using smartphone apps; information (about bank-specific vulnerabilities) can now travel at unprecedented speed via social media; and the banking sector is being affected by unusually high volatility in the interest rate cycle.13
To date, the recent consolidation wave looks comparable to previous waves, when scaled relative to U.S. GDP, as shown in the chart below. Prior waves included the savings and loan crisis in the late 1980s to early 1990s, and the GFC in 2008–09.
Although the acute phase of the current U.S. banking crisis appears to be coming to an end, additional bank failures cannot be ruled out entirely. In the coming quarters and years, market participants may continue to question the viability of certain business models in an environment of tighter regulation and higher funding costs.14
How does AI fit into the shortages and excesses framework?
The AI revolution is currently visible as an expectation in financial markets; that is, it is not yet impacting real economic data. Simply put, if AI were driving the real economy, inflation would be among the least of economists’ worries, because the high productivity gains in production would quickly alleviate shortages in the real economy.
That said, the capital markets are pricing in an inflection point after which the adoption of AI may profoundly affect corporate earnings. This approach lets us fit a consideration of AI into our shortages/excesses framework: The outperformance of AI-related stocks—which are up 37% year to date through July—can be seen as driven by a perceived shortage of the necessary infrastructure required to power the AI revolution.
Have we seen this before?
In a word, yes. The most recent example would be the cloud computing revolution, which started with the launch of Amazon Web Services (AWS) in 2006. By 2011–2012, AWS was reporting accelerating revenue growth and had no real competition; Google and Microsoft Corporation didn’t begin to accrue meaningful gains in their cloud-based businesses’ revenue until 2015 or later.
At the time, the market assigned AWS's parent company, Amazon.com, a 70x–80x forward P/E valuation in recognition of its new technology (which, thanks to economies of scale, promised to substantially reduce data service costs for corporations). Amazon also benefited from a significant first-mover advantage: AWS had a competitive “moat,” or gap, surrounding its cloud-based business, which boosted the parent company’s earnings outlook.
Amazon’s multiple may have looked like an overvaluation at the time, but—as we look back now—it clearly wasn’t. The market was simply pricing in a dominant winner at the forefront of a business revolution sparked by a new technology.15
Nvidia currently trades at a 50x forward P/E multiple, which is lower than Amazon’s comparable multiple back in 2012–2015. Moreover, Nvidia’s competitive advantage in designing the chips used for generative AI is arguably even bigger than Amazon’s historical edge in cloud computing. GPU chip design entails laborious research and requires decades of internal company knowledge and patent creation; cloud computing may be more capital-intensive, but it does not require the same intensive research and development.17, 18
How do investors perceive Nvidia? Q1 2023 was a watershed moment. The company beat already lofty earnings estimates by more than 30%, signaling that accelerated AI-related demand from data centers is real and likely to continue.
Here’s another analogy: After Apple released its first iPhone toward the end of 2007, the company’s subsequent earnings growth soared—and yet market participants failed to appreciate the product’s full implications. In every quarter from 2009–2012, Apple consistently beat earnings estimates by more than 20%. During this period, iPhone revenue rose from USD 13 billion annually to USD 79 billion. (A decade later, in 2022, iPhone revenue came in at USD 205 billion.)19 In 2009, Apple traded at a 30x forward price-earnings ratio, and just like Amazon in 2012–2015, that multiple wasn’t an overvaluation.
Many market observers anticipated that 2023 would be a year driven by macro factors, including the possibility—at the start of the year—of recession in the United States. That hasn’t happened, but large sectoral disparities have emerged in U.S. capital markets. Using a framework that emphasizes the role of shortages vs. excesses in the real economy can help investors make sense of these wildly different sector performances.
Although AI promises to be an exciting new investment theme, is the market getting ahead of itself when it comes to this groundbreaking technology’s potential benefits? We think the answer is “probably not,” based on our analysis of historical trading patterns and valuation changes in the wake of earlier technological breakthroughs (cloud computing and the iPhone).
AI certainly holds the promise of driving labor productivity growth in the years ahead. Yet the most exciting opportunity, in our view, will come from the intersection of AI and the real economy: Could AI help alleviate the shortages that have accumulated in the real economy from years (and in some cases decades) of underinvestment? This applies not only to housing and autos, where shortages have persisted post-pandemic, but also to the U.S. energy system and the nation’s infrastructure, which appears destined to benefit from a renewed focus on U.S. industrial policy. (See our recent article on renewed U.S. industrial policy.20)
In the months to come, investors will need to remain flexible and alert to the potential that shifting patterns of demand and supply may impact sectors of the economy differently. New shortages may appear and fundamental supply excesses may also be revealed. Further sector-specific rolling recessions may be inevitable if interest rates continue to rise, but some industries will likely keep benefitting from robust and ongoing demand. The four sectors we have identified—not to mention the enormous potential impact of AI on all corporate activity—provide some indication of just how careful and selective investors need to be.
1.This is not to say that U.S. recession risks are currently low; they continue to be elevated relative to historical norms and will remain so until there is more clarity surrounding potential job losses as the labor market continues to rebalance amid restrictive monetary policy. Also, when we say the economy is running too hot, we are referring to the inflation process rather than real economic activity. Real economic activity has been running at a below-potential pace of around 1% over roughly the last year. See: Federal Reserve Bank of New York, Weekly Economic Index, June, 2023.
2High yield spreads over U.S. Treasuries have narrowed by approximately 70 basis points YTD as of mid-July 2023.
3Auto sector returns are heavily skewed by the doubling of Tesla’s share price this YTD through June 23, 2023. Excluding Tesla, the S&P 1500 Auto sector is still up 18% YTD June 23, 2023, which is above broad market returns.
4Global X Funds' Global X Artificial Intelligence & Technology ETF.
5Joe Seydl, Dan Alter, “Looking to buy or sell a house in today’s strange U.S. market?” J.P. Morgan Private Bank, April 17, 2023. More specifically, we assume the housing market was in equilibrium in 1996, and then we consider the difference between the average of starts/completions and trend household formation from 1996 to present, assuming 200,000 housing units lost each year due to obsolescence.
6The weighted average cost of raw materials used to produce an automobile rose 116% YoY in 2021, the most ever. Raw material costs have since been subsiding but remain materially elevated vs. pre-pandemic, and non-commodity supply chain costs remain at a premium as well. In addition, labor costs will likely take another step up with UAW contracts up for renewal this September.
7Ryan Brinkman & Rajat Gupta, “2023 CMD Fleshes Out Margin Expansion Strategy; Work Cut Out for Model But We Think Pro Can Deliver” J.P. Morgan Securities Research, May 23, 2023.
8Another way of looking at the valuation of Tesla would be to consider its forward price-to-earnings-to-growth (PEG) ratio, which—at 1.96—is below that of other fast-growing companies, such as Apple (2.22) and Microsoft (2.04).
9We define the forward P/E discount as the discount relative to the average differential vs. the S&P 500 Index over the past 10 years.
10Joe Seydl, Jay Serpe, Ryan Asato, “Are banks vulnerable to a crisis in commercial real estate?” J.P. Morgan Private Bank, April 12, 2023.
11We say “as a result of the pandemic shock” because there are other areas of CRE that are distressed, such as older retail shopping malls due to the penetration of online shopping. The retail distress is not new, however, and has not been the primary focus of investors in recent quarters.
12The United States has far more banks than many other countries due to the history of its banking system. Until 1994, various financial regulations prevented banks from operating on a national scale—or even across state lines. The passage of the Riegle-Neal Interstate Banking Act of 1994, however, finally lifted all interstate banking restrictions. Since then, the number of U.S. banks has decreased by a rate of ~3% annually. (Sources: Marc Rubinstein, “The US Is Paying the Price for Being Overbanked,” Bloomberg, March 29, 2023 and “Why America has so many banks,” The Economist, May 26, 2023.)
13The rates cycle is important, but it needs to be understood in the context of accelerating deposit outflows at some unusual institutions that lacked a “sticky” deposit base. There is no inherent reason why the banking system as a whole should be especially sensitive to higher interest rates, given that deposit franchise values tend to rise to offset asset value losses when interest rates rise. See: Itamar Drechsler, Alexi Savov, and Philipp Schnabl, “Banking on Deposits: Maturity Transformation without Interest Rate Risk”, The Journal of Finance, June 2021.
14From a regulatory perspective, increased capital and funding requirements are likely. We also note growing bipartisan political support for lowering the global systemically important banks (GSIB) regulatory threshold from USD 250 billion to USD 100 billion (but that change is not likely to be enacted until after the 2024 U.S. presidential election).
15Another way to look at this would involve considering the stock performance of Amazon after its peak forward P/E multiple was achieved in February 2015. Since then until now, the total return of Amazon’s stock has reached 582%, which compares to 152% for the S&P 500 Index as a whole.
16Unlike past forms of AI, generative AI produces novel, human-like output in the form of text, images and three-dimensional models, and in some contexts has been shown to be able to match or exceed human benchmarks.
17Kunjan Sobhani and Oscar Hernandez Tejada, “Expanding Nvidia's Total Addressable Market,” Bloomberg Intelligence, May 22, 2023.
18Nvidia’s competitive advantage can also be attributed to the company’s AI-related software offerings (in addition to chip design). For example, Nvidia provides a suite of cloud services, pre-trained AI models, optimized inference engines, and application programming interfaces (APIs) for specific enterprise applications, including those in the domains of robotics, automotives and the health care industry.
19David Curry, “Apple Statistics 2023,” Business of Apps, May 2, 2023.
20Joe Seydl, Jessica Matthews, Ian Schaeffer, “The opportunity in renewed U.S. industrial policy,” J.P. Morgan Private Bank, June 1, 2023.