Although not a board member, Faust was arguably the third most powerful figure at the Fed after the chairman and William C. Dudley, president of the Federal Reserve Bank of New York. The fact that he was unknown outside the Fed only enhanced the power of his hidden hand, used to move markets with a few choice words. In spy novel parlance, Faust was the “man without a face.”
Wordsmithing forward guidance is not done from the sidelines. Providing precise phrases for use in public disclosures required intimate knowledge of FOMC inner workings, and the private views of Bernanke and Yellen. Faust sat in the Fed’s ornate, high-ceilinged boardroom at almost every FOMC meeting during his time there, which included the Fed’s largest program of money printing called QE3, Bernanke’s threat to reduce money printing in a May 2013 speech, and the actual reduction in money printing beginning in December 2013. Between FOMC meetings, Faust was in Bernanke’s and Yellen’s offices for the brainstorming sessions as words and phrases were tried out for potential impact on markets. When I later spoke to Bernanke about Faust’s role, he told me, “Yes, Jon’s office was just across the hall from mine.” Concerning forward guidance, Faust was the Fed’s brain.
Faust is a member in good standing of the Keynesian-monetarist academic coterie. He received his Ph.D. in 1988 from the University of California, Berkeley. Janet Yellen was a professor at Berkeley before she became a senior Fed official. Faust’s thesis adviser, the Nobel Prize–winning economist George Akerlof, is married to Yellen. Faust worked at the Fed in various capacities from 1981 to 2006, ultimately becoming assistant director in the international finance division. Suffice to say, Faust was no stranger to the Fed, Bernanke, or Yellen when he got a call to advise the board in 2012.
On January 20, 2015, not long after he left the Fed, Jon and I had dinner at The National, a popular New York steakhouse, in a private dining room on the second floor. The National’s décor is typical for the cuisine—dark wood, brass trim, white tablecloths, and dim lighting. I had known Jon for years. Still, this was our first chance to meet in person since Bernanke tapped him in 2012. We sat at right angles, a foot apart. Our conversation continued for two hours over salmon, crème brûlée, and good wine; Jon drank red, and I my usual sauvignon blanc.
In addition to keeping up with speeches and writings by FOMC members, I spoke occasionally with Fed governors and reserve bank presidents, so I had some reference points on Faust’s FOMC colleagues. That made the conversation livelier because we could share impressions on personalities in addition to policy.
I was especially interested in Jeremy Stein. Stein was a governor and FOMC member from May 2012 to May 2014, dates that overlapped with Jon’s time at the Fed. Stein struck me as the only governor with a technical grasp of how the Fed’s zero interest rate policy was creating hidden dangers and bubble dynamics. Some FOMC members at the time, including Dallas Fed president Richard Fisher, were outspoken about the need to raise rates and the dangers of not doing so. Yet Fisher, and the like-minded Charles Plosser of the Philadelphia Fed, had intuitive, even populist reasons for raising rates. These reasons had in part to do with the unfairness of not giving depositors a decent return on their money while Wall Street bankers used easy money to enrich themselves with leveraged stock buybacks.
Stein was subtler. He saw inside the machine. Stein knew that asset swaps—an exchange of junk collateral for good collateral so the exchanging party could pledge good collateral in another deal—were adding hidden leverage. He understood that increased regulation was driving disintermediation—so-called shadow banking—making it worse than what collapsed in 2008. He grasped that derivatives risk was in gross notional values, not net. This was clear from his speeches and writings. Stein saw the bubble dynamics. Then he was gone. No one left on the FOMC seemed to see what Stein saw.
My question to Jon was straightforward. Stein had sounded a warning inside the Fed. His analysis was rigorous, not populist. Stein also knew that if a new bubble burst so soon after 2008, that would destroy confidence for a generation and undo the work the Fed had done since the last crisis to move the economy to a self-sustaining path. I leaned forward and asked Jon, “Does Yellen see what Stein saw? Does she believe markets are in a bubble?” Then came Faust’s reply: “Not yet.”
This reply was revealing. It meant the Fed was sticking to obsolete models. The idea that the Fed should not try to pop bubbles, but instead clean up the mess after they pop, has a long pedigree. Discussion of this approach goes back at least as far as the classic work of Friedman and Schwartz on the origins of the Great Depression. Friedman and Schwartz were critical of the Fed’s decision to raise interest rates in 1928 to cool off a stock market bubble. By raising rates at a time when inflation was not a threat, the Fed induced a recession in 1929, which was a proximate cause of the stock market crash in October of that year. That crash is frequently cited as marking the onset of the Great Depression. Both Alan Greenspan and Ben Bernanke support the Friedman and Schwartz critique. Greenspan received praise for letting the dot-com bubble that began in 1996 pop on its own in 2000. Greenspan “cleaned up the mess” without serious economic damage or systemic contagion. Bernanke echoed Greenspan’s approach to handling bubbles in extensive writings and in a landmark speech on the causes of the Great Depression given on March 2, 2004.
Yet, the Greenspan-Bernanke approach to bubbles is both a misreading of history and contradicted by more recent experience. The Fed did blunder by raising rates in 1928, but the blunder was not that they attacked a bubble, but that they failed to follow the rules of the game. The United States was on a gold standard in 1928 and saw extensive gold inflows from Europe. Under the monetary rules of the game, gold inflows required monetary ease that would cause inflation, raise export prices, and rebalance gold flows in favor of Europe. Raising rates increased gold inflows to the United States and reduced liquidity in the rest of the world. This policy was the opposite of that required under a gold standard, and was a direct contributor to the Great Depression.
What Greenspan and Bernanke miss is that there is neither a gold standard today, nor any monetary standard at all. Without a monetary anchor to gauge policy, the Fed must think harder about whether it is the cause of a bubble rather than a mere bystander. Decisions to raise or lower rates are not guided by gold inflows, but rather by whims, and spurious correlations between inflation and employment known as NAIRU (the non-accelerating inflation rate of unemployment) and the Phillips curve.
Experience shows that Greenspan’s handling of the dot-com bubble was not so deft after all. His cleanup included keeping interest rates too low for too long, which led directly to the housing bubble and the 2008 financial collapse. Bernanke’s zero interest rate policy (continued by Yellen) from 2008 to 2015 repeated Greenspan’s mistake with catastrophic potential.
The better analysis is that bubbles are not automatically dangerous. What matters is whether they are fueled by debt or not. The dot-com bubble was inflated by what Greenspan earlier called “irrational exuberance,” not debt, and did relatively little harm to the macroeconomy when it burst despite investor losses. In contrast, the mortgage bubble was driven entirely by debt and derivatives and caused the greatest recession since the Great Depression. All bubbles are not alike. Stein understood this.
In understanding these bubble dynamics, leverage is a more apt metric than debt. Leverage can include derivatives in addition to traditional borrowings. This was Stein’s other great insight. Bernanke and Yellen not only failed to distinguish between debt- versus non-debt-driven bubbles, but failed to see that derivatives were a form of debt. The new asset bubbles still expanding in 2016 were the bad type—driven by debt and derivatives. Yellen’s obsolete hands-off approach missed this distinction.
There’s nothing new about economists’ inability to foresee panics. One infamous example was prominent economist Irving Fisher’s observation that stock markets had reached “what looks like a permanently high plateau,�
� made days before the October 28, 1929, stock market crash that took the market down 24 percent in two days. That crash continued as the stock market fell 80 percent from the 1929 high before bottoming out in 1932. The point is not to ridicule Fisher—he was one of the twentieth century’s most brilliant economists—but rather to make the point that economists, especially those at the Fed, simply don’t see bubbles.
There are models that do a good job identifying bubbles using complexity theory, causal inference, and behavioral economics, although the exact timing of collapse remains difficult to predict due to the minuteness of catalysts, and the stochastics of path dependence. Jeremy Stein and former Fed governor Rick Mishkin have made the most progress in the use of recursive functions to grasp these risks. Still, Faust’s reply put a spike in my hope that this thinking was more than a novelty at the Fed. In Yellen’s mind, it was business as usual. There was no bubble.
The other disturbing feature of Faust’s reply was the word “yet.” This carried an implication that while a bubble might be in the making, there was still time to keep it under control. The notion is that it is possible for central bankers to let the air out of a balloon slowly. Another metaphor is a thermostat. If a house is too cold, you can dial the thermostat up. If the house is too warm, you dial it down. The connotation is that controlling the economy is a linear and reversible process. All it takes is a turn of the dial.
An economy more nearly resembles a nuclear reactor than a thermostat. Reactors can also be dialed up or down. Yet that process is neither linear nor necessarily reversible. When a supercritical state is reached, a reactor core melts down. No amount of dialing can cause a melt-up. Nuclear reactors are complex systems, as are capital markets. Faust was unintentionally saying the Fed has no idea how capital markets behave.
In contrast to his comment on bubbles, Faust’s remarks on quantitative easing were refreshing. He candidly acknowledged that inside the Fed, QE’s effect was considered “murky.” QE seemed worth doing, and was probably better than nothing, but any beneficial impact was unclear.
Bernanke admitted as much when we spoke in 2015. I heard the same in private conversations with other FOMC members. They admitted they did not know what they were doing after 2008. Bernanke told me his ideal was FDR during the Great Depression. FDR was a great improviser. His administration would pursue a policy on not much more than a hunch it might work. Some ideas did work, some failed. Others had some impact, yet were blatantly illegal and later overruled by the courts. It didn’t matter. FDR’s mantra was to try all possible paths out of economic distress. FDR felt that in a crisis it was better to do something than nothing. Bernanke told me he agreed with that approach.
In fact, it depends. At times, it is better to do nothing than to flail about. That is the essence of medicine’s Hippocratic oath, which in modern English reads, “I will not be ashamed to say ‘I know not.’ . . . Above all, I must not play at God,” and “prevention is preferable to cure.” The historical record strongly suggests the Great Depression would have ended sooner but for uncertainty caused by FDR’s improvisations. The prolonged depression since 2008 (defined as persistent below-trend growth) owes to uncertainty caused by Bernanke-Yellen improvisations. In a state of regime uncertainty, capital goes on strike.
Faust was equally candid about the drafting sessions for the press releases issued after each FOMC meeting. He called the process “ridiculous.” Significant word changes were made for the sake of change with no intrinsic meaning attached to the words. The reader was led on a hermeneutic semiotic voyage. Bernanke once looked at two drafts of an FOMC statement; one meant to signal no policy change, and one intended to signal tightening. Bernanke looked up and asked Jon, “Which one’s the bad one?” The words were arbitrary, and for show. Faust’s more important job was to call Jon Hilsenrath, a reporter for The Wall Street Journal, to explain what the Fed wanted the words to mean regardless of word choices. Hilsenrath dutifully reported the intended meaning, and the market responded as expected. Michel Foucault would have been proud.
After about an hour of conversation, Faust and I pivoted from policy issues to discuss the post-2008 era in historical context. Notwithstanding my Fed critique, I allowed that the Fed at least thought it was pursuing the right strategies. Yet this raised a counterfactual: If the Fed knew then what it knows now, would it have followed the same path? The premise behind the question was simple. When the Fed launched QE2 in November 2010, it expected robust results by 2011. When it launched QE3 in September 2012, it expected robust results by 2013. The hoped-for growth had not materialized. The economy was not worse; jobs were created; yet growth was weak, well below potential. Was the whole process a box canyon, a dead end from which there was now no exit?
Faust didn’t answer the question directly. Instead he mused that fifty years from now, “a new Ben Bernanke will come along, a young scholar who would look back at the 1930s and this period we’re in now and compare the approaches to see what worked and what didn’t. That person will have two data points.” Faust’s dry academic remark about data points opened another window on Fed thinking.
In addition to obsolete equilibrium models, Fed staff adhere to what are called frequentist statistical methods. The frequentist method stands in contrast to another statistical style, an inferential approach based on Bayes’ theorem. Both approaches, frequentist and Bayesian, wrestle with cause and effect to make forecasts. Like many dichotomous debates, there has been some synthesis in recent years; users of each approach can see something salutary in the other. But battle lines are still drawn.
Frequentists assert that statistically valid conclusions can be drawn only when large data sets with long time series are available; the larger and longer, the better. These large data sets are sorted and analyzed through baselines, regressions, and correlations to hypothesize causality and spot anomalies. That output forms a foundation for robust predictions of future behavior. One particular technique used in economics is the Monte Carlo simulation. Computers are used to spin a simulated roulette wheel or roll digital dice millions of times with output plotted into degree distributions so specific outcome frequencies can be observed and forecast with confidence. The more data, and more frequent the observation, the more confident the statistician is in her forecast—thus the name “frequentist.”
Bayesians work with far less data, not because they want to, but because sometimes one must. If you need to solve a life-or-death problem, and have only one data point, Bayes’ theorem helps you find an answer. Bayesians solve problems when data sets are small, even nonexistent. They do so by making an assumption, called a prior, and use the prior to form a hypothesis. The prior is based on factors including history, common sense, intuition, and what scant data might exist. The prior is assigned some probability of being true based on the best available evidence. In the absence of any data, the prior is assigned a fifty-fifty probability, the best approximation of uncertainty.
The Bayesian reasons backward from subsequent observations to test the prior hypothesis. Each subsequent event is evaluated for the separate probability that it would or would not occur if the hypothesis is true. Then the prior is updated to increase or decrease its probability of being correct. Over time the prior can become quite strong, with a 90 percent probability of being correct; or it can become weak, in which case it is discarded. The good Bayesian keeps an open mind about the impact of subsequent observations on the original hypothesis. Frequentists are horrified at the guesswork in developing the prior and assigning probabilities in the absence of data. They view the method as nonscientific, only slightly better than voodoo.
Bayesians rebut frequentists with pragmatism—what if you don’t have a lot of data, yet the problem can’t wait? What if U-boats have cut off Britain’s food supply, and your mission is to crack the Nazi Enigma code and break the U-boat blockade? By the time frequentists had enough data to solve the problem, the United Kingdom would have been star
ved into surrender. This is why Bayesian approaches are widely used in military and intelligence operations. Those operators face existential questions that cannot wait for more data.
Faust’s remark was frequentist to the bone. In effect, he said the Bernanke-Yellen policies from 2007 to 2015 had only one data point to process—the Great Depression. Bernanke ranks as a monetary scholar of the Great Depression just behind Milton Friedman and Anna Schwartz, giants of that field. Fifty years from now, presumably in the midst of another economic misadventure, a future policymaker would have two depressions to study, 1929 and 2008, and could perhaps compare and contrast policy responses as if answering a question on a college final exam. In fairness to Faust, Bernanke expressed the same view when I spoke to him. The former chairman ruminated that it was too soon to tell if his policies had been successful. It would take future scholarship, decades hence, to render that judgment.
My assessment of the Bernanke-Faust frequentist mind-set was that at the rate of one data point per century, we’d be well on the way to understanding linkages between monetary policy and depressions by the year 2525. What I took Faust to mean was that when crisis struck in 2008, Bernanke had only one frame of reference, and did the best he could. Interestingly, Janet Yellen’s academic outpost between her stints as a central banker was the University of California at Berkeley, the intellectual ground zero of frequentist statistical science for the past century. Yellen was even more data driven and model based than Bernanke.
The world’s misfortune was that the Fed had no firm grasp of Bayesian technique. Capital markets were condemned to a succession of calamities while academics-turned-central-bankers waited decades for more data to convince them of their failures.
The Road to Ruin Page 20
