I have been asked many times about whether and how the Federal Reserve considers asset prices (such as stock prices and house prices) in its determination of the appropriate level of interest rates.

Specifically, some people suggest that the Fed should raise interest rates when asset values appear high relative to historical norms to stop asset bubbles from forming, such as the tech bubble in the late 1990s and the housing bubble in the mid-2000s. After all, when the housing bubble burst, it was devastating for the economy, causing the financial crisis and the Great Recession. Wouldn’t the economy have been better off if the Fed had simply raised rates when the bubble first started forming and thus avoided all that harm?

The purpose of this essay is to explain how I think about Federal Reserve policies to address potential bubbles. This topic seems simple, but I will argue it is highly complex, with large potential consequences for Main Street. Let me remind readers that my comments are my own, and do not necessarily represent the views of the Federal Reserve System.

In summary, I will explain five points: (1) It is really hard to spot bubbles with any confidence before they burst. (2) The Fed has limited policy tools to stop a bubble from growing, even if we thought we spotted one. (3) The costs of making policy mistakes can be very high, so we must proceed with caution. (4) What we can and must do is ensure that the financial system is strong enough to withstand the inevitable bursting of a bubble. And finally (5) monetary policy should be used only as a last resort to address asset prices, because the costs to the economy of such a policy response are potentially so large.

The Federal Reserve’s mandates

In 1977, Congress gave the Fed its dual mandate: stable prices and maximum employment. However, we can’t ignore the implicit role the Fed also has to try to achieve financial stability. After all, when Congress first created the Fed in 1913, it did so in response to financial crises that repeatedly hammered the U.S. economy in the late 1800s and in the panic of 1907. The Board of Governors and 12 regional Federal Reserve Banks were specifically created with the goal of promoting financial stability. Price stability and maximum employment came almost 70 years later.

Achieving financial stability is hard — really hard. Human societies are prone to mass delusion and to bubbles; history has numerous examples, from the tulip bubble in Holland in the 1600s to the stock market bubble in the 1920s to the housing bubble in the 2000s. Future generations are exceptionally good at repeating past mistakes. Even if we focus just on the Fed’s official dual mandate, financial crises can cause very high unemployment and low inflation or even deflation. My perspective is that whether it is officially acknowledged or not, whether we want the responsibility or not, the Fed has an important role to try to ensure financial stability. So where does monetary policy fit in?

Spotting bubbles is hard

Everyone can recognize a bubble after it bursts, and then many people convince themselves that they saw it on the way up. Michael Lewis’ highly entertaining book, “The Big Short,” is a perfect example. With the benefit of hindsight, Lewis picked four guys who happened to be right this time. It would have been far more impressive if Lewis had identified and written about them when the housing bubble was forming. Why didn’t he? Because on any given day, there are lots of people predicting various doomsday scenarios (spend a little time on finance Twitter to see for yourself). How do you know which one is right among all the cranks? And maybe today’s crank will look brilliant tomorrow. As they say about broken clocks…

I will offer three examples of well-intentioned government officials trying unsuccessfully to accurately identify bubbles and potential crises.

1. In 1996, then-Fed Chairman Alan Greenspan gave his famous “irrational exuberance” speech, where he said the stock market was overvalued. Essentially, Greenspan was warning that a bubble was forming and that investors needed to be careful because a correction was coming. At the time, the S&P 500 had a price-to-earnings ratio of 17.8. The stock market did end up correcting in the early 2000s, after the P/E ratio reached 26.9 by the end of 1999. The P/E ratio fell below 16 by mid-2002. Was Greenspan right when he called a bubble in 1996? Should the Fed have raised interest rates in response? Of course, it is impossible to know what would have happened if Greenspan had used monetary policy to act on his irrational exuberance call. But given how the stock market has climbed in the following 20 years, I would say that this was a “false positive” — identifying a bubble far too early, or seeing one where it didn’t exist.

2. When I went to Treasury in July 2006, then-Treasury Secretary Henry Paulson declared to his staff that the U.S. economy was due for some form of crisis. He didn’t know where it would come from but, because markets had been stable for some time, history suggested something would happen. So he tasked his staff (including me) to work with the Federal Reserve and Securities and Exchange Commission to look for signs of trouble. We looked at a variety of scenarios, from an individual large bank running into trouble to a hedge fund blowing up. Sadly (and embarrassingly), we never considered a nationwide housing downturn. We missed it, and we were looking. It seems obvious now. This was clearly a “false negative.”

3. Finally, in the wake of the 2008 financial crisis, all regulators were on alert for potential economic shocks. We had all learned our lessons and weren’t going to make the same mistake again. Yet I don’t know anyone who predicted oil prices climbing to over $100 per barrel and then falling to $26. That is an enormous price decline, and we all missed it. Was oil at $100 a bubble? Should the Fed have raised interest rates when oil started its climb? I’m not sure if oil was a bubble or not.

The Fed’s policy tools to slow down asset price increases are limited

The Fed’s primary policy tool is setting short-term interest rates. When inflation is lower than our 2% target and unemployment is high, we lower interest rates to try to stimulate economic activity by reducing borrowing costs. When inflation is high and unemployment low, we raise rates to try to prevent the economy from overheating.

Monetary policy is a blunt instrument: We set the overnight interest rate, and it then affects rates all across the country, across different asset classes. That’s one of the biggest challenges in trying to use monetary policy to change asset prices. For example, if we see a bubble forming in commercial real estate, raising interest rates won’t affect just the commercial real estate market, but also housing, automobiles, consumer borrowing and capital-intensive industries, among others. We may want consumers to keep spending, but condo prices to stop rising. Raising interest rates would slow them both down.

The Fed also has regulatory and supervisory tools that it can use to change the behavior of the financial institutions it supervises. For example, the Fed regularly issues and explains its expectations for banks in the form of letters. The letters become guideposts that our examiners use to ensure that firms are safe. But the letters can also influence asset markets by changing bank actions.

In 2013, regulatory agencies, including the Fed, observed that lending to companies that were leveraged was rising rapidly and that interest rates on leveraged loans were falling. In response, we issued guidance on leveraged financing to remind the industry of particular risk management expectations and to warn banks holding large amounts of these loans that they would be subject to additional regulatory scrutiny. This letter was followed up by supervisory actions by bank regulators that had the effect of penalizing banks that did not follow the regulations. By ensuring that banks lend appropriately in this market, this supervisory activity could affect the price of high-yield bonds and related investments.

However, the Fed does not have all of the targeted tools to address individual asset markets that some other countries have. While the Fed can limit the amount of debt used to buy stocks, some countries can also adjust the loan-to-value (LTV) requirements of mortgages. By increasing the down payment requirement (lowering LTVs), those countries could directly target the housing market if it were showing signs of overheating. This is an example of a highly targeted tool that, in theory, should be effective in slowing down the housing market without slowing down the entire economy the way raising rates would.

Even with additional tools, bubbles can be exceptionally difficult to slow down

Looking at other countries’ experiences in trying to deal with potential asset bubbles shows how difficult it can be to slow down rapid increases in asset prices. Regulators in Sweden and Canada have tried to use more powerful and targeted tools than the Federal Reserve has to cool their housing markets. As powerful as those tools appear to be in theory, they have not been very effective in slowing down price appreciation.

• Given fast increases in housing prices in Sweden, in October 2010, the Finansinspektionen (FSA) applied an LTV limit of 85% to any new mortgage or extensions to existing mortgages that used a home as collateral. Prior to this policy, the average LTV ratio on new loans was over 70%, with more than 33% of these loans having an LTV ratio above 85%. In November 2014, the FSA introduced a mandatory amortization of mortgages for those with an LTV ratio higher than 50%. Home prices in Sweden have continued to increase despite these policy actions.

• Authorities in Vancouver have been concerned about increasing home prices for a number of years. The price appreciation seems to be driven by foreign buyers who are looking for safe, offshore investments, which has made buying homes unaffordable for many Vancouver residents. In August 2016, local authorities passed a 15% tax on home purchases by foreign buyers. This was a very targeted and, in theory, very powerful policy tool to curb further price appreciation. Price growth in Vancouver did slow for a time, but appears to be climbing again. And prices in Toronto, which does not have such a tax, now seem to be climbing even faster.

My takeaway from these countries’ experiences is that when asset prices are climbing rapidly, they can be very difficult to slow down, even with policy tools that are targeted squarely at the asset class. That suggests to me that if central bankers were to try to use monetary policy to slow those bubbles down, the rate increases necessary to be effective would likely be large, resulting in high economic cost to the rest of the economy.

The costs of false positives can be very high

Imagine if the Greenspan Fed had decided to use monetary policy beginning in 1996 to stop stock prices from climbing further. How high would interest rates have had to go? What would the economic costs have been? I don’t know for sure, but it seems possible that the Fed would have had to push the economy into a recession to stop stock prices from rising further.

Similarly, imagine if the Fed had identified oil as a potential bubble when it started its climb after the Great Recession. How high would the Fed have had to raise rates to stop oil prices from rising further? What would the costs have been?

Given how hard it is to slow down price increases when a bubble is forming (as discussed earlier), I assume the monetary policy response would need to be large enough to risk putting the economy into recession to stop a bubble. So I ask myself: If we think we might see a bubble, are we confident enough that it is worth putting the economy into recession to stop it? Hence, the bar must be high before we should consider using monetary policy to address asset prices.

The costs of false negatives are sometimes very high — but not always

The housing bust, financial crisis and resulting Great Recession were devastating for the American people. Millions lost their jobs, their homes and their savings. It has been a frustratingly slow recovery, and I believe it has directly led to the deep political divisions in the country that we are still experiencing.

But not all asset busts are so costly. When the tech bubble burst in 2000, equity investors lost money, but it led to only a mild, fairly short recession. It seems to me that the Fed was right to not try to slow down equity markets in the late 1990s. The cost of prevention by raising interest rates may have exceeded the cost of the correction. Similarly, I mentioned oil price spikes and falls in the last decade. No doubt these swings were painful for the oil sector, including North Dakota, which is in my Federal Reserve District, but the costs to the economy overall have been small. As with the tech bubble, had the Fed tried to use interest rates to prevent oil prices from rising, the cost of prevention would likely have exceeded the cost of the correction.

My takeaway from the varied costs of false negatives is that we must first try to assess the cost of a correction before we determine whether to try to address asset prices that appear elevated.

Debt seems to be the key risk in bubbles

What determines if an asset price correction will trigger a crisis or just a milder downturn? Debt seems to be a key factor. Housing is a huge, highly leveraged market. The mortgage market is roughly a $10 trillion market. Today, people often buy homes with 20% as a down payment. Going into the financial crisis, people were putting little to nothing down with those infamous no-doc loans. Those loans were bundled into mortgage-backed securities, which were then bundled into collateralized debt obligations, and then banks bought them with yet more borrowed money. It was leverage on top of leverage with little equity supporting it all.

Contrast that with the stock market, where individual investors are limited to a 50% margin. In other words, those investors have to put at least 50% down on their equity investments. Most put down much more. Similarly, mutual funds tend to have much less leverage than mortgage lenders, for example. Direct comparisons between the leverage underlying the housing market and the stock market is complex, but I believe it is safe to say that as an asset class, housing is far, far more leveraged than the stock market.

When the stock market corrects, investors lose money. Technology companies with no profits go bankrupt. But the economy as a whole does not seem as vulnerable as when a large, highly leveraged asset class such as housing corrects. The Fed’s job is not to protect investors. It is to promote financial stability. Sometimes those overlap. Not always.

Public awareness matters when considering potential policy responses

The Fed’s job is “to take away the punch bowl just as the party gets going,” as former Fed Chairman William McChesney Martin famously quipped in 1955. In other words, it’s the Fed’s job to put the brakes on the economy before it overheats, which almost by definition will be unpopular with many people. So why should public awareness matter to the Fed?

Regulators don’t exist in a vacuum. The Fed ultimately gets its power from the American people, through authorities granted to it by their elected representatives. Yes, the Fed must make tough, sometimes unpopular choices, but the ability of the Fed to impose and sustain steep costs on the economy and Main Street is limited by the willingness of the people to accept those costs.

The most famous example of the Fed imposing steep costs on Main Street for the long-term health of the economy is the Volcker Fed dramatically raising interest rates to crush inflation in the early 1980s. The costs were large — a deep recession and unemployment reaching 10%, the same as the peak unemployment rate in the Great Recession. But a key factor that enabled the Fed to impose such painful medicine was that the American people hated inflation. In public opinion surveys, inflation was ranked as the number one economic issue on voters’ minds. So when Chairman Volcker explained that high rates and the resulting economic costs were necessary to control inflation, while the public was still angry with the Fed, they at least agreed on the problem.

Now imagine a potential bubble that only the Fed sees. Imagine if Chairman Greenspan declared war on housing prices in 2004, when many Americans were enjoying homeownership for the first time. I suspect many Americans, members of Congress, homebuilders, realtors, banks (and many others) would have been outraged that the Fed was depriving people of participating in the American dream for a problem that they didn’t believe was real. Imagine if the Fed decided to raise rates to prevent housing prices from climbing. Given how painful the Great Recession was, this may have in fact been a better choice than what the Fed actually did, which was basically nothing. But I have doubts about whether the Fed could have maintained the policy long enough for it to have the desired effect before the American people rejected it.¹ At a minimum, the Fed would have had to work very hard to try to convince the public that the housing bubble was real and a danger. It can be very hard for a sole regulator to stand up against a national belief that home prices only go up and say: “We know better than all of you.” But that doesn’t mean we shouldn’t try.

The Fed has stronger tools to mitigate the damage from bubbles than to prevent them

This essay has been pretty skeptical about the powers of the Fed to identify and slow down bubbles. But there is something we can do that does not require us to identify bubbles in the first place. We can make sure our financial institutions are sound and can withstand the shock of asset price corrections. Without question, one of the key factors that magnified the intensity and costs of the 2008 financial crisis was the undercapitalization of the nation’s largest banks. They amplified the shock rather than dampened it. If we make sure the largest banks are highly capitalized (the Minneapolis Fed’s estimate is that they need roughly double the equity capital they currently have), the financial system will be much more resilient against asset price corrections in the future.

In addition, if we identified an asset class that appeared richly valued to which banks had a lot of exposure, we could use existing tools to respond. This is the essence of the current stress test. The Fed tests how a decline in asset values in a weak economy would affect the solvency of a bank. And banks that have insufficient capital to withstand such losses can face reductions to their dividends and share buybacks to ensure that they build adequate capital to withstand a correction. The Fed could potentially use other existing tools to accomplish the same goals or ask for new tools as necessary.

Conclusion: What does all this mean? A strategy for the Fed

OK, so how do I put this all together? First, I hope I have convinced you that identifying bubbles on the way up is extremely difficult, and it will be rare indeed when we make such an identification with any confidence. What we should do now is make sure the financial system is resilient and can withstand a future correction. That means we should force the large banks to raise a lot more capital now, when markets are strong.

Second, given how costly some asset bubbles can be when they burst, even though it is difficult, we must remain on alert, always looking for signs of a new bubble forming. If, however unlikely, we do spot a potential bubble, then we must try to assess how damaging a correction would be.

If we think it’s not likely to be very damaging, then we shouldn’t do anything, because the cost of false positives is high. However, in those cases where debt is fueling the asset value increase, a correction could trigger financial instability, because banks might take huge losses and potentially fail.

The worry is not the high asset values themselves, but the exposure of market participants to those assets. In those cases, we should consider a number of options: (1) Speak out to raise awareness of the potential bubble, but not just one mention of “irrational exuberance.” If we really think we see an iceberg ahead, we should be speaking out until people take the risk seriously. (2) Use what nonmonetary tools we have to try to make sure the financial system is positioned to withstand the coming correction (by limiting bank dividends, for example). (3) Ask Congress for new authorities to make the financial system more resilient, which admittedly might take too long to be useful. (4) I would say as a last resort, if we are confident that the potential bubble poses grave danger, consider raising interest rates to try to slow it down.

Keep in mind, by the time we are confident that a dangerous bubble is indeed forming, it may be too late. Raising rates aggressively at that point might just burst the bubble, causing the very harm we hope to avoid.

Given the challenges of identifying bubbles with any confidence and the costs of making a policy mistake, I believe the odds of circumstances ever making sense to use monetary policy to try to slow asset prices down are very low. I won’t say never — but a whole lot of evidence would have to line up just right for it to be the prudent course of action.

Addendum: What might be wrong with my analysis

I always ask myself what my analysis might be missing. One argument not yet captured in this piece is that low rates, which might be appropriate given where inflation and unemployment are, could make bubbles more likely to form in the first place. Much has been written about the low neutral real rate environment we are currently in and expect to be in for the foreseeable future. Basically, due to a range of macroeconomic factors (such as demographic trends and low productivity growth), the interest rate that is neutral, i.e., that neither stimulates nor restrains the economy, is lower than it has been in recent decades. Current estimates are that the neutral real rate (net of inflation) is currently around zero or perhaps slightly negative. Could it be that such low rates make bubbles more likely to form and, if so, what should we do about it?

The truth is we don’t have a good answer to this question. If inflation is low and there is slack in the labor market, how high should we raise rates to reduce the chances of bubbles forming? We don’t have a good economic theory to analyze this scenario and offer policy guidance. It is a question that needs more research. Until we have such a theory that we have confidence in, I believe we should continue to focus on our dual mandate goals to set monetary policy and then keep our eyes open for potential bubbles and respond as best we can. The cost of keeping rates high to reduce the chances for future bubbles would be higher unemployment and a risk of unanchoring inflation expectations to the downside. Those are large economic costs.

Endnote

1) The banking agencies did issue proposed guidance in January 2006 to banks to have them better manage the risk of commercial real estate concentrations, which included single-family home developments. The agencies received about 4,500 comment letters in response, expressing what they called “strong opposition to the proposed guidance.” As a result, the agencies reported that in the final guidance, they “revised the proposal to clarify that financial institutions play a vital role in providing credit for commercial real estate activity and to make clear that the Guidance does not establish a limit on an institution’s CRE lending activity.” See https://www.gpo.gov/fdsys/pkg/FR-2006-12-12/html/06-9630.htm.

Neel Kashkari is president of the Federal Reserve Bank of Minneapolis.