I recently posted a detailed article in response to Raj Chetty’s lament that scientists make fun of economics’ pretense to science.
The thrust of my article was that the problem was not that economics was inherently incapable of becoming more scientific. The problem was that so many economists wear ideological blinders that recurrently cause them to perform a parody of the scientific method.
Chetty claimed that economists who are “testing precise hypotheses” in quantitative studies that exploit natural experiments are (finally, in 2013) “transforming economics into a field firmly grounded in fact.” Chetty’s metaphor is that economics is like epidemiology. (One assumes that his column is posted in the CDC’s common areas for the general amusement of epidemiologists.)
“As is the case with epidemiologists, the fundamental challenge faced by economists — and a root cause of many disagreements in the field — is our limited ability to run experiments.
(Surely we don’t want to create more financial crises just to understand how they work.)
Nonetheless, economists have recently begun to overcome these challenges by developing tools that approximate scientific experiments to obtain compelling answers to specific policy questions.”
My column pointed out that there was nothing new about economists studying natural experiments and that the methodology produced a field “firmly grounded in [fiction]” rather than “fact.” I went through a series of “scientific” economic studies during a financial crisis that led to policy recommendations by economists conducting the studies that were exceptionally criminogenic. The economists consistently praised the worst frauds and damned our reregulation of the industry, which proved successful in stopping the S&L fraud epidemic. I showed that the field’s top economists specializing in finance consistently got things as wrong as it is possible to get it wrong because they did not understand fraud or fraud mechanisms and relied on “data” that were the product of accounting control fraud. In virtually every case, the economists who purported to study natural experiments by “testing precise hypotheses” implicitly excluded control fraud as a possible explanatory variable. I explained the primitive tribal taboo that economists have that prevents them from using the “f” word and noted that economists do not study fraud or fraud schemes. Economists implicitly exclude control fraud, in contravention of the scientific method, for ideological reasons that they frequently do not even consciously recognize. Economists that study finance wear blinders that are invisible to them.
Every day provides multiple examples of the blinders imposed by economists’ dogmas reducing their purportedly scientific studies about finance to exercises in self-parody. John Taylor provide a classic example in his October 28, 2013 op ed in the Wall Street Journal.
Taylor begins with a strong claim as to the causality of the financial crisis.
“The crisis did not reflect some inherent defect of the market system that needed to be corrected, as many Americans have been led to believe. Rather it grew out of faulty government policies.”
Taylor is one of the most famous hard-right neo-classical economists, so his twin claims about the market system and government provide a concrete example by an economist from the highest echelon of the field that allows us to test Chetty’s claim that top economists embrace the scientific method regardless of their ideologies.
Chetty rightly reminds us that science must be grounded in facts, so the key question is: did Taylor use the scientific method to prove his twin claims of causality? The answer is “no.” We know that he could not do so because the kind of study he relies on for proof of causality is inherently incapable of determining causality. I guarantee the reader that Taylor has made that point to students hundreds of times during his career. Here is one example in which Taylor was discussing the causes of the financial crisis and warning against the “confusion of correlation with causation.”
It is a point we all make, frequently, in teaching economics. Taylor has violated the most basic of empirical rules – conflating correlation with causation.
Taylor’s assertion of causation fails the most minimal adherence to the scientific method.
“[E]xtremely low interest rates led individual and institutional investors to search for yield and to engage in excessive risk taking, as Geert Bekaert of Columbia University and his colleagues showed in a study published by the European Central Bank in July.”
Bekaert’s study fails to comport with the scientific method because it too conflates correlation with causality. Bekaert concedes that despite four prior econometric studies of the issue “no extant research establishes a firm empirical link between monetary policy and risk aversion in asset markets.” The conclusion reiterates: “there is no empirical evidence on the links between risk aversion in financial markets and monetary policy.” Note that that the ECB authors are explaining that prior studies have not even demonstrated correlation.
At some points in the description of their study, Bekaert and his colleagues play it straight by describing a correlation and noting that some writers have asserted causality.
“The strong interaction between the VIX index, known as a “fear index” (Whaley (2000)), and monetary policy indicators may have important implications for a number of literatures. First, the recent crisis has rekindled the idea that loose monetary policy may lead to excessive risk-taking in financial markets.”
Within a three paragraphs, however, the authors abandon the scientific method and assert causality. “A lax monetary policy decreases risk aversion in the stock market after about nine months. This effect is persistent, lasting for more than two years.” Why would “lax monetary policy” (a term the authors do not expressly define) have such a large lag before it purportedly drove risk aversion? Why would it cease to drive risk aversion after two years? The authors offer no clues, which is remarkable given their claims of causality. There is no logical or theoretical basis for their purported lag and persistence. Under the “efficient markets” and “rational expectations” theories that are essential to the ECB authors’ study the markets should immediately take into account all public information. Indeed, under rational expectations theory, the markets should anticipate that Federal Reserve under Alan Greenspan would push down rates and incorporate the Fed’s practice into current prices months, even years, before Greenspan lowered rates. To the extent that the authors suggest any theory to explain the lags and persistence, the logical implications of their theory are that there should be no lag and the effect of “lax monetary policy” should be permanent.
Pages later, the authors implicitly admit that their study cannot establish causality because “the relationship between risk aversion and monetary policy may also reflect the joint response to an omitted variable….” Yes, there are an enormous number of variable that could actually determine causality and it could be the interaction effects of many omitted variables that is most important in determining if there is any reliable “relationship” between risk aversion and monetary policy. Because of these omitted variables Bekaert’s study inherently cannot establish even reliable correlation between risk aversion and monetary policy because he did not control for these variables or their interaction. Correlation cannot establish causation and controlling for some variables does not change that basic proposition of the scientific method. I return to this point and provide specifics below.
But there are more basic problems with using the VIX index than conflating correlation with causation. The ECB authors rely on a two-step process to purportedly derive their measure of risk aversion from the VIX index. They state that the VIX index purports to be a measure of “the stock market option-based implied volatility.” The authors claim that they can derive separate measures of S&P 500 option investors’ uncertainty and risk aversion by decomposing the VIX index.
There are multiple, disabling problems with using the VIX index to measure the risk aversion of originators and purchasers of mortgages and mortgage derivatives. First, no one knows whether the VIX index actually measures the implied volatility of the stock market (or, more precisely, the portion of the stock market that the VIX index purportedly relates to). The VIX index was constructed on a theoretical basis with the theory arising from the efficient market hypothesis. The theory is that if a portion of the stock market were efficient an investor could buy an option to invest in that would hedge the “implied volatility” in those stocks. In order to operationalize this theory the creators of the VIX index had to specify how it would be calculated by making a series of choices about the date and nature of the stocks and precise options to use. These calculations are approximations that would introduce errors even if the VIX were theoretically sound, particularly because the composition of the subset of stocks (the S&P 500) used to calculate the VIX and define the option used as the hedge changes over time as corporations fail or grow or shrink significantly.
It is highly unlikely that the theoretical basis for constructing the VIX index is sound. Markets were grotesquely inefficient during the financial crisis – the relevant time period. The claim that derivatives had a consistent relationship to underlying assets such as the stock market also proved false during the crisis.
The result of these uncertainties is that we do not know whether the VIX index ever accurately represented the “implied volatility” of the stock market. We do not know whether the VIX index’s asserted relationship to the implied volatility of the stock market is consistent or varies. There was never any way to prove whether the VIX index was accurate and represented a consistent relationship with stock market volatility. The VIX is a faith-based index premised on a theory that proved false.
Second, there are strong reasons to fear that the so-called fear index is becoming an ever poorer measure of stock market implied volatility.
“Mike Pringle, global head of equity trading at [Citi], told Reuters that the VIX volatility index , is now as much a traded asset as it is a guide to investors seeking protection from losses.
The VIX reflects Standard & Poor’s 500 .SPX options prices and, therefore, expectations of future market moves. The idea is that as people become fearful of losing their money, they are more willing to buy a put option as protection.
At the moment, it remains at very low levels.
‘A big mistake the market makes is looking at the VIX as an indicator of stock market risk. Why? Because it’s an asset class and it’s more traded for yield than protection,’ Pringle said.
‘The growth of structured products around VIX drove that move. In most cases, the VIX is sold to generate yield but during some stress periods, the weakness in the spot level triggers significant computer-generated technical buying from these products,’ he said.
The VIX is widely followed as an indicator of investor sentiment, although there has long been debate over its efficacy as more financial instruments are derived from it.
‘It’s still relevant in extremes, but not in a normal functioning market,’ Pringle said.
Pringle cited Citi trading strategy research showing tens of billions of dollars’ worth of assets under management linked to the VIX through structured notes, which had to be rebalanced to reflect actual market moves. This dampened volatility, he said.
The Citi data showed these VIX-related contracts make up about 34 percent of overall volatility trading on the S&P 500 and as much as 44 percent of the short-term, 2-month volatility.
The VIX rose nearly 80 percent in three weeks after the Fed hinted in late May at ‘tapering’ its stimulus, albeit from a low starting point. It has since given back almost half the gains.
At levels around 15, it is currently below its average of just over 20, nearly half the 2012 high of 28 and well off the 2011 high of 48 and all-time high of nearly 90.
COMPUTERS DRIVING VOLUMES
While persuading others of the VIX’s flaws is not easy, Pringle said Citi’s handling of risk management in equities had been restructured accordingly.
Rather than relying solely on the VIX, Citi traders and clients can turn to the Central Risk Desk, through which a large proportion of its trades are routed.
The computer programs that underpin the desk’s activities assess around 60 measures of market stress and timing – from global risk arbitrage spreads to dividends to repo rates – to get a better read on sentiment, behavior and deal timing.
Looked through this prism, there is greater risk currently in global markets than the narrower VIX is suggesting.
‘Basically every ‘fear factor’ you could possibly get your hands on,’ Pringle said. Such a move had helped the bank be more efficient and grab 2 percent of market share in Europe, the Middle East and Africa last year, he said.”
Both aspects of these developments render the ECB authors’ use of the VIX index unreliable as a variable. One of the empirical problems that arises when there are strong reasons to fear that its relationship to “implied volatility” has grown is weaker is obvious. A weak relationship renders the index useless. But an index that changes over time in its relationship to “implied volatility” causes even greater empirical unreliability if the study looks at any material time period (and the ECB study does).
Third, there is, therefore, no way of proving whether the techniques the authors used to purportedly decompose the index and obtain a measure of risk aversion by stock market investors actually measure their aversion. The authors’ methodology is inherently a faith-based asserted decomposition of a faith-based VIX index. We cannot reliably test whether either the VIX index or the decomposition is correct. As I have explained, the theoretical basis the authors present of their faith in the index and their asserted decomposition has been falsified. It is far more likely, therefore, that the decomposition techniques add error to an index that was constructed on the basis of a false theory. We cannot measure either error because we cannot measure overall risk (conceived of as implied volatility) or either assumed subset of implied volatility (“risk aversion” and “uncertainty”).
The ECB authors admit to a telling bit of uncertainty about “uncertainty” in their discussion of purportedly decomposing the VIX index.
“However, the VIX index, used by Bloom (2009) to measure uncertainty, can be decomposed into a component that reflects actual expected stock market volatility (uncertainty) and a residual, the so-called variance premium, that reflects risk aversion and other non-linear pricing effects, perhaps even Knightian uncertainty. Establishing which component drives the strong co-movements between the monetary policy stance and the VIX is therefore particularly important.”
The ECB authors admit that it is “particularly important” that there be a reliable means of decomposing “risk aversion” from the VIX index. Recall that the starting point – the authors’ claim that the VIX index accurately and consistently measures the “implied volatility” in S&P 500 options is unproven, unprovable, and likely to be false. Compounding that fundamental problem, it turns out that the authors do not have even a faith-based theoretical basis for their purported means of decomposing the VIX index in order to separate out “risk aversion.” The additional problem is that “risk aversion” is under their (failed) theory part of a “residual” along with other “non-linear pricing effects” – and the authors do not have even a failed theoretical basis for knowing what other variables are included in that residual. They think “perhaps” that it might “even [include] Knightian uncertainty.” They are fatally uncertain about uncertainty even if one assumed that their faith in their “Vixen theories” were justified. The ECB authors’ decomposition methodology, therefore, is unproven, unprovable, uncertain, and likely to be false because it is based on failed and incomplete theories. One cannot ascribe a residual to a single variable (risk aversion) when one knows that the residual has multiple contributors and is unsure what those contributors are or how they would even theoretically perform. They cannot decompose when they do not know the composition or have a means of identifying the subcomponents. Their use of the word “perhaps” is a devastating clue that the ECB authors knew they had no reliable means of decomposing already faith-based VIX index.
Fourth, there is no way to know whether the risk aversion of those who buy S&P 500 stock options tracks the risk aversion of bankers making fraudulent real estate loans or investment bankers purchasing such fraudulent loans in the secondary market. They are different people purchasing radically different assets. The stock option purchasers are passive investors – they do not expect to run the corporation or determine its policies. The CEOs that lead accounting control frauds run the business. They care about their wealth – not the wealth of shareholders. They loot the shareholders and use the shareholders’ capital to suborn and the lobbyists who maximize the three “de’s” (deregulation, desupervision, and de facto deregulation) that allow the CEO the ability to loot with impunity.
The CEOs’ abuse of their power to optimize looting produces corporate investments in high risk assets with underwriting practices so pathetic that they create intense adverse selection and a negative expected value to the lending without competently underwriting. Neo-classical economists erroneously interpret the combination of the lenders’ asset choices and the destruction of underwriting as evidence of falling risk averseness, but the reality is that the CEO is risk averse and uses his power and the fraud “recipe” to loot with impunity. I discuss this in more detail immediately below when I return to explaining the specific differences between causation and correlation in the context of risk aversion in the home mortgage context. The ECB authors do not even attempt to explain why the purchasers of S&P 500 options would share the same risk averseness as the CEOs of the mortgage lenders and purchasers that led the looting.
The VIX index, the authors’ key variable, cannot be used reliably to study even correlation, much less causality, much less causality in the mortgage industry fraud context. Microeconomic variables that the ECB authors did not study can drive the VIX index. Consider Panels A & B of their paper. They claim that the decomposed VIX data demonstrate that the period 2004-2007 represents a period in which investors exhibited “low uncertainty” and “high risk appetite.” The first claim is true of the CEOs leading the fraud epidemics, but the second is false for the most important sector that drove the crisis – the housing mortgage markets.
The ECB authors do not cite George Akerlof and Paul Romer’s famous 1993 article – “Looting: The Economic Underworld of Bankruptcy for Profit.” They do not cite the criminology literature on accounting control fraud. They do not discuss the fact that epidemics of accounting control fraud drove the second phase of the S&L debacle and the Enron-era scandals. They do not discuss the three fraud epidemics that drove the current financial crisis. The twin mortgage origination fraud epidemics were inflated appraisals and endemically fraudulent liar’s loans. Fraudulently originated loans can only be sold through fraudulent “reps and warranties,” so the twin origination fraud epidemics led to the epidemic of fraudulent sales of the fraudulently originated mortgages to the secondary market. Individually, each of these epidemics of accounting control fraud would have constituted the most destructive financial frauds in world history. Collectively, they hyper-inflated the financial bubble and caused the catastrophic losses that drove the financial crisis and the Great Recession.
Akerlof and Romer emphasized the key analytical points that the ECB authors missed when thy implicitly assumed accounting control fraud out of existence.
“[M]any economists still [do] not understand that a combination of circumstances in the 1980s made it very easy to loot a [bank] with little risk of prosecution. Once this is clear, it becomes obvious that high-risk strategies that would pay off only in some states of the world were only for the timid. Why abuse the system to pursue a gamble that might pay off when you can exploit a sure thing with little risk of prosecution?”
[S]omeone who is gambling that his thrift might actually make a profit would never operate the way many thrifts did, with total disregard for even the most basic principles of lending: maintaining reasonable documentation about loans, protecting against external fraud and abuse, verifying information on loan applications, even bothering to have borrowers fill out loan applications” (Akerlof & Romer 1993: 4-5).
As I have explained, Akerlof and Romer also warned about the perverse incentives of loan brokers and sellers of mortgages to the secondary market and the capacity of epidemics of control fraud to hyper-inflate financial bubbles. Akerlof and Romer provided the key that would have allowed any “scientific” economist to unlock the crisis. They anticipated in uncanny ways the key drivers of the current crisis. Akerlof was awarded the Riksbank Prize in Economics in 2001 (economics’ version of the Nobel Prize). No “scientific” economist could fail to discuss Akerlof and Romer and “looting” as a possible cause of a financial crisis. In particular, no scientific economist trying to explain the origination and purchase of seemingly ultra-high-risk mortgages and mortgage derivatives could fail to discuss looting as a potential cause. The ECB economists and Taylor, however, are the overwhelming norm among economists. They implicitly exclude the existence of accounting control fraud. They do not cite the relevant economics or criminology literature.
Accounting control fraud epidemics produce a “sure thing” so “uncertainty” is minimized. Accounting control fraud fools economists when it comes to risk and risk aversion. The assets that the frauds invest in are higher risk in order to obtain an (apparent) premium yield, but the “sure thing” remains so, from the prospective of the officers running the fraud, “risk” is minimal. The officers are risk averse – the purpose of the fraud scheme is to produce huge wealth for the officers with certainty. The officers leading the mortgage origination and secondary market frauds took enormous steps to make the fraudulent mortgage assets appear exceptionally low risk. Liar’s loans dramatically inflated the borrowers’ income while endemic appraisal fraud grossly understated the loan-to-value (LTV) ratio. The CEOs leading the frauds created the Gresham’s dynamics that suborned purported “independent” “professional” “controls” into opinions that aided the fraud, including faux due diligence by entities like Clayton, “clean opinions” from top tier audit firms, grossly inflated appraisals, and “AAA” ratings by the three top rating agencies.
Recall Chetty’s claim that economists are scientists analogous to epidemiologists. As a criminologist, I love the metaphor. It is one we use to great effect in our field. We borrowed the concept of a pathogenic environment that can produce epidemics. We search for and warn against criminogenic environment that can produce fraud epidemics. We look for “vectors” that spread fraud epidemics. We study how fraud transmits and how frauds seek to suborn the analog of the body’s immune system. We “autopsy” failures to learn these facts and develop our theories and the “natural experiments” we study empirically. With the exceptions of economists like Akerlof and Romer, however, economists overwhelmingly do not act like epidemiologists when it comes to control fraud epidemics. Economists “know” that financial markets “self-correct” and promptly eliminate fraud. Fraud epidemics do not and cannot exist. In preparing to study this crisis, an epidemiologist would have known that we had suffered two modern financial driven by fraud epidemics. No one questions that fact that the Enron-era scandals were accounting control frauds. The national commission to investigate the causes of the S&L debacle reported.
“The typical large failure [grew] at an extremely rapid rate, achieving high concentrations of assets in risky ventures…. [E]very accounting trick available was used…. Evidence of fraud was invariably present as was the ability of the operators to ‘milk’ the organization” (NCFIRRE 1993).
An epidemiologist would have been particularly interested in the non-crisis in this era. This was the 1990-1991surge in lending that we now call liar’s loans that we drove out of the S&L industry. Recall Akerlof and Romer’s warnings about loans made without underwriting being strong evidence of accounting control fraud.
Given the dominant role of accounting control fraud in the S&L debacle, the 1990-1991 genesis of liar’s loans, and the Enron-era scandals an epidemiologist would have known where to start looking if she were asked to study this crisis. She would have begun by looking for evidence of accounting control fraud.
Economists act like epidemiologists would act in response to an epidemic if epidemiologists were members of a religion that had a taboo about discussing or studying disease and believed that the body self-corrects if left alone by physicians. Epidemiologists who believed that epidemics do not and cannot exist would never look for the presence of bacteria or viruses, never look for pathogenic environments, and never study the autopsy results. What would we call an epidemiologist who did not believe that epidemics can exist, did not study what made environments pathogenic, did not study bacteria or viruses, and did not study autopsy results? We could call them a theoclassical economist. They would, implicitly, ignore fraud. They would propose anti-regulatory policies that were intensely criminogenic.
Similarly, what would we call an engineer who designed homes that had an average life expectancy of three years before they collapsed and caused enormous losses? I’d call them a “financial engineer.” Bad economists are very good at appropriating these favorable professional labels to try to pretend to science. Anyone familiar with how a real engineer is trained and their professional ethos of safety and care for protecting others from injury knows that Wall Street “financial engineers” are the opposite of real engineers. Wall Street “financial engineers” are selected on the basis of their ethos of making themselves wealthy at the expense of their clients and then mocking the clients when they suffer losses. Financial engineers view customers as sheep to be sheared (the actual Wall Street verb for what should be done to customers are too vulgar for publication).
Only a theoclassical economist like John Taylor would proclaim, without ever studying or discussing the three most destructive fraud epidemics in world history, would issue a dogmatic statement that this is the first virgin crisis, conceived without sin the corporate “C” suites.
“The crisis did not reflect some inherent defect of the market system that needed to be corrected, as many Americans have been led to believe. Rather it grew out of faulty government policies.”
The VIX index can be driven by many variables. I believe it is increased appreciation of the extent and nature of risk, particularly fraud risk, rather than risk aversion changes, that is the primary driver of extreme changes in the VIX index. I do not believe that even if there were a reliable VIX S&P 500 stock option index and even if it could be reliably decomposed to quantify stock aversion changes among the purchasers of those options that the index would explain the behavior of the CEOs that led the three massive epidemics of accounting control fraud that hyper-inflated the bubble and drove the crisis. There is no evidence that these CEOs decisions were the product of changes in the CEOs’ risk aversion, but the relevant means by which the CEOs leading the frauds changed risk was by using a fraud scheme that greatly reduced the risk of detection and sanction and by using their political power to maximize the three “de’s.” The rate of growth of the money supply (Taylor’s fixation) has nothing to do with the risk aversion of the CEOs that led the three fraud epidemics. Neither Taylor nor the ECB authors understand the relevant nature of the CEOs’ risk aversion or how the CEOs that led the three epic control fraud epidemics minimized their risk while greatly increasing the firm’s risk of loss.