“… positive trends can sometimes evolve into prices that increase more than fundamentals justify.” Eric Rosengren, President and CEO, Federal Reserve Bank of Boston, May 9, 2017.
The Federal Reserve’s annual stress test is the de facto capital planning regime for the largest U.S. banks. Not surprisingly, it comes under frequent attack from bank CEOs who argue, as Jamie Dimon recently did, that “banks have too much capital…and more of that capital can be safely used to finance the economy” (see page 22 here). From their perspective, this makes sense. Bank shareholders, who the CEOs represent, benefit from the upside in good times, but do not bear the full costs when the financial system falters. As readers of this blog know, we’ve argued frequently that capital requirements should be raised further in order to better align banks’ private incentives with those of society (see, for example, here and here).
A more compelling criticism of central bank stress tests focuses on their discretionary character. To the extent feasible, central banks should minimize their interference in the allocation of resources by private intermediaries, allowing them to direct lending to those projects deemed to be the most productive. For this reason, frequent alteration of asset risk weights (used in calculating a bank’s risk-weighted capital) can be counterproductive, leading to herding and unnecessary asset price volatility. Similarly, frequent fine tuning of stress test scenarios may do little to make the financial system safer if banks have no time to respond.
But the painful lessons that have come from large asset price fluctuations and high concentrations of risk provide a strong case for the kind of limited discretion that the Fed uses in formulating its stress tests. This year’s test—in which the severely adverse scenario assumes a 35-percent plunge in commercial real estate (CRE) prices—is a useful case in point. U.S. CRE prices have nearly doubled since their post-crisis trough in early 2010. Even adjusted for inflation, they reached a record at the end of 2016, up by more than 75 percent from the trough (see chart).
Real CRE Price Index (2010=100), 1972-4Q 2016
This blog post highlights why it makes sense for regulators to use the stress test exercise to learn how well the largest U.S. intermediaries would fare if the recent CRE boom were to turn into a bust. The point is that, from time to time, having policy discretion to alter the most severe scenarios is a useful way to assess and promote resilience of the financial system.
For a variety of reasons, the CRE sector is particularly important for the health of many financial systems. First, it is frequently large. In the United States, for example, the current market value of real estate owned by nonfinancial corporations accounts for nearly one-third of their total assets―more than $13 trillion (see table B.103 in the Financial Accounts of the United States). Second, as the chart above indicates, CRE prices exhibit very sizable and persistent swings. Third, and perhaps most important, since it is a common source of collateral, leveraged intermediaries like banks often have significant exposure to CRE.
There is no shortage of historical episodes in which CRE prices exhibit large, persistent swings. One of the most dramatic is Japan’s boom and bust that helped usher in an era of zombie banks and zombie borrowers. As the next chart shows, in the decade to 1991, land prices used for commercial properties in Japan’s largest cities rose six-fold. Then, over the succeeding decade, they plunged by more than 80 percent. Even now, 25 years later, prices remain more than 80 percent below their 1991 peak! It is no wonder that credit supply first surged and then collapsed, driven by (and amplifying) land price movements.
Japan: Index of Urban Land Prices for Commercial Property (2000=100), 1972-2016
Had Japan’s regulators employed a 21st-century-style stress testing and capital planning regime in the latter half of the 1980s, they might well have lessened both the asset price swings and the long-run damage.
So, why does all this matter today in the United States? It turns out that there is substantial evidence of froth in US CRE prices. To see what we mean, consider the prices for all-equity real estate investment trusts (REITs), which serve as a reasonably liquid market proxy for CRE prices. Since 2010, the REIT price-dividend ratio has averaged 27.7, or more than double the norm (13.2) of the three decades through 2003, before the run-up to the financial crisis (see the next chart). In theory, a high price-dividend ratio could reflect either low expected returns in the future or expectations of rapid dividend growth. It is tempting to think that the low interest rates and bond yields that have prevailed since 2010 are consistent with the former.
U.S. All-Equity REIT Price-Dividend ratio, 1972-April 2017
However, important new research by van Nieuwerburgh (Why Are REITS Currently So Expensive?) argues compellingly otherwise. Using multifactor asset pricing models, van Nieuwerburgh decomposes the high-price dividend ratio into its two parts. He then concludes that a rising risk premium has offset the favorable impact from the low safe-asset (Treasury) interest rate. As a result, the expected returns on CRE assets are not unusually low. Rather, the model-based estimates of one-year-ahead dividend growth expectations are unusually high, ranging between 20 and 30 percent annually for most of the period since 2010. This is roughly five times the 5.6 percent average over the period since 1972 (and double the 12.8 percent average after 2000). Such optimism about dividend growth makes real estate prices “more vulnerable than ever to an increase in interest rates and/or an economic contraction” (see page 38 here).
U.S. bank regulators have been concerned about CRE exposure for years. For example, even before the financial crisis, the three federal banking regulators warned about concentrations in CRE lending (see here). In 2015, the same group issued an “interagency statement on prudent risk management for CRE lending.” More recently, in the financial stability section of its February 2017 Monetary Policy Report to Congress, the Federal Reserve expressed concern about CRE valuations, an issue that it had raised in each of the previous four semi-annual reports beginning in February 2015.
In a recent speech, President Rosengren of the Federal Reserve Bank of Boston identified various factors that could cause optimistic CRE valuations to give way. For example, any combination of tighter monetary policy, a pickup in global growth, and an upside inflation surprise would drive up long-term interest rates—which are what matter for CRE. In addition, reform of the GSEs, which play a leading role in the multifamily mortgage market, could boost the financing risk premium in the apartment sector. Indeed, the capitalization rate (the ratio of net operating income to price) for multifamily dwellings hit new lows this year and is well below comparable rates in the industrial, office, and retail segments of the CRE universe (see Figures 1, 13 and 14 in President Rosengren’s slides).
It is important to keep in mind that CRE prices matter for financial stability. At the end of 2016, CRE loans (including commercial mortgages and multifamily residential mortgages) totaled $3.8 trillion, half of which are owned by U.S.-chartered depository institutions. To put this into perspective, depositories’ exposure to CRE is slightly higher than the $1.8 trillion net worth of commercial banks. And exposure is growing, with holdings up by 9 percent over the previous year.
This brings us back to this year’s stress test. The 35-percent plunge of CRE prices included in the 2017 scenario is anything but arbitrary. First, between the final quarter of 2007 and the first quarter of 2010, CRE prices plunged by 40 percent. Second, the recent price run-up is comparable in scale to the one preceding the 2007-2010 crash. Third, as indicated, policymakers have been warning about CRE valuations for years, so banks have had plenty of time to prepare. Fourth, the severely adverse scenarios in 2015 and 2016 already included a 30 percent drop in CRE prices. So, if anything, this year’s change results in a decline that is too small.
It is worth emphasizing how important it is that authorities retain some degree of discretion in formulating the stress scenarios. Doing otherwise courts catastrophe. The best evidence for this is the way in which stress tests were carried out on the U.S. government-sponsored enterprises (GSEs) in the run-up to the financial crisis. Unfortunately, as Frame, Gerardi and Willen note, the tests were static, with the model specification and parameters fixed, and the stress scenario “insufficiently dire.” The predictable result was that the tests under-predicted GSE losses (and capital needs) that materialized when house prices plunged. When the GSEs failed, taxpayers bore the losses.
The ultimate difficulty is that no one, neither regulators nor private market participants, can know whether and to what extent asset prices are deviating from levels warranted by fundamentals. Put differently, we cannot identify bubbles that make the financial system vulnerable, let alone know when they will burst. This is the challenge highlighted by the initial quote from President Rosengren. And, as Minneapolis Fed President Neel Kashkari recently argued, it is an important element in the case for higher capital requirements: we need the largest banks to be resilient in the face of potentially large, and inherently unpredictable, asset price collapses.
It also is a key reason why policymakers need some leeway to adjust their stress tests to changing economic and financial conditions. With bank capital, as with so many aspects of the financial system, it is far better to be safe than sorry.