Marc D. Hauser was a prominent professor in the Psychology Department at Harvard, renowned for his work in animal and human cognition. Hauser stood out on campus, with his trendy goatee, mischievous eyes, and perfect-posture demeanor, a man who clearly enjoyed Alpha-male status. He was a lauded public intellectual, well known among students, fellow scientists, and the popular press. Then, in 2010, the news leaked that Harvard had found Hauser responsible for eight counts of unspecified scientific misconduct. A year later he resigned his position at the university. While neither Hauser nor Harvard has fully specified what he did wrong, the journal Cognition retracted a 2002 paper by Hauser, and media reports suggested that Harvard decided to investigate Hauser’s lab after students who had worked there made allegations of data falsification.1
According to the Chronicle of Higher Education, former research assistants of Hauser became suspicious of his research and eventually reported their concerns to Harvard administrators. Hauser had long argued that primates (specifically rhesus monkeys and cotton-top tamarins) could recognize patterns as well as human infants could. The research assistants accused him of falsely coding videotapes of monkey behavior.2 For one experiment, the Chronicle wrote, Hauser’s team watched videotapes and coded how the monkeys reacted. If the monkeys recognized the pattern, it was coded one way; if they failed to, it was coded another. Standard research protocol called for two researchers to independently watch and code a videotape to confirm the reliability of the coding. In the experiment profiled by the Chronicle, Hauser and a research assistant both watched and coded the tape. Another research assistant analyzed the results and found that the first assistant had not observed the monkeys seeing the change in pattern that was predicted by Hauser’s research. However, Hauser’s coding showed that the monkeys had noticed the change in pattern. If Hauser’s coding was correct, the experiment was a success and publishable; if the research assistant’s coding was correct, the experiment was a failure. The second research assistant and a graduate student in the lab suggested to Hauser that a third researcher should code the results, but Hauser argued against this proposal. After several rounds of emails with his junior colleagues, Hauser wrote, “I am getting a bit pissed here.” The research assistant who analyzed the data and the graduate student reviewed the tapes themselves, without Hauser’s permission, and coded the results independently. They each concluded that the experiment had failed and that Hauser’s coding did not reflect the behavior of the monkeys in the video. As word about this incident spread, several other lab members noted that they had had conflicts with Hauser in which they believed he reported false data and insisted on its use.
In 2010, just before news of Hauser’s misconduct broke, I gave a presentation to a conference of senior officials of the Dutch government in The Hague. The aim of the conference was to share the insights of behavioral decision researchers and psychologists with government policymakers. In the United States most of the social science that makes it to policymakers comes from the realm of economics, but Dutch officials appeared to be seriously interested in what psychology had to offer them. Diederik A. Stapel, a prominent psychology professor from Tilburg University in the Netherlands, was a central figure in the effort to bring psychological research to the Dutch government’s attention and was part of the team that invited me to The Hague. Unfortunately, soon after the conference Stapel was suspended from Tilburg University for fabricating data over the course of years, affecting at least thirty of his publications. On October 31, 2011, a Tilburg University committee revealed that three unidentified junior colleagues of Stapel’s had initially reported his fraudulent behavior. In the Hauser and Stapel stories, it required multiple junior colleagues sharing their concerns with one another to build the support necessary to go public with evidence against their senior colleague.
Responding to the interim report, Stapel stated:
I failed as a scientist. I adapted research data and fabricated research. Not once, but several times, not for a short period, but over a longer period of time. I realize that I have shocked and angered my colleagues because of my behavior. I put my field, social psychology, in a bad light. I am ashamed of it, and I deeply regret it. . . . I did not withstand the pressure to score, to publish, the pressure to get better over time. I wanted too much, too fast. In a system where there are few checks and balances, where people work alone, I took the wrong turn. I want to emphasize that the mistakes that I made were not born out of selfish ends.3
Even more recently two highly visible psychologists have been added to the list of scholars who are alleged to have created fraudulent data. All of these stories of data fraud are tragic, and they were covered thoroughly by the media. It is interesting that in both the Hauser and Stapel cases, a few junior researchers noticed what was occurring and acted on their observations, while many other, more senior colleagues who had access to the same information did nothing. It bears stating that, first, I wish the punishment for data fraud was more severe. Second, such cases are a minor threat to scientific progress. Fraudulent researchers are rare, and the scientific norm of replicating important results provides significant protection from false conclusions. When a researcher reports results that the scholarly community cannot later replicate, his or her work is generally viewed with suspicion.
The rare case of outright fraud exposed can blind us to a greater concern. I am not referring to the possibility of systemic outright fraud that fails to get exposed. As shocking as these two cases are, the fact is that junior colleagues did share their suspicions and did eventually act. There is the far greater risk of a field, or an industry, taking a collective sigh of relief when outrageous transgressions of integrity and ethics are brought to light. The majority can confidently declare that they are not guilty of anything similar. But noticing the outright fraud can blind us to a threat to integrity far more pervasive than fraud and far less visible to the media—a threat that can blind entire industries. What happens when accepted practices are insufficient to secure an industry’s claim to integrity?
In March 2012 my excellent colleague Don Moore and I were invited to testify before the U.S. Public Company Accounting Oversight Board (PCAOB), created by the Sarbanes-Oxley Act of 2002 in response to the economic failures of Enron, its auditor Arthur Andersen, and a host of other companies. The PCAOB was holding hearings on the question of whether organizations should be required to regularly change audit firms in order to help create auditor independence. Our invitation to the PCAOB’s offices in Washington, D.C. was somewhat an act of theater. James Doty, the chair of the PCAOB, already knew what we would say, since we had been repeating the same argument for the past fifteen years—and doing a lousy job of convincing institutional leaders to create true auditor independence. Sitting in front of the five commissioners and an audience made up of auditors opposed to reform and the business media, we enthusiastically repeated our arguments for auditor independence one more time.
This discussion is sounding pretty bureaucratic, so let me offer a bit of background. The U.S. government, like the governments of most developed economies, recognizes that many external parties (investors, strategic partners, etc.) need to be able to rely on a company’s accounting books when making decisions about investing and interacting with that company. As a result, corporations are required to be audited by independent audit firms, and an industry was created specifically for the purpose of providing independent audits. In 1984, writing on behalf of a unanimous U.S. Supreme Court in the case of United States v. Arthur Young & Company, Chief Justice Warren Burger argued that auditors were required to “maintain total independence from the client at all times.”
In my testimony to the PCAOB, I made the same argument that I had made initially in a 1997 article and then in testimony before the Securities and Exchange Commission in 2000, and that I will present again here for you: namely, that we do not have independent audits in this country, that the steps needed to create independence are amazingly clear, and that we continue to fail to make these changes. The United States has a costly requirement that firms be independently audited, yet the industry was established in such a way that the one thing it cannot provide is independence. And the “Final 4” accounting firms that dominate this industry have been enormously successful at manipulating the U.S. legal and political systems to maintain their markets and profits at the expense of those for whom the requirement of independent audits was created. Meanwhile most segments of society have failed to notice the magnitude of this problem.
So, what’s the problem, exactly? It is that we have institutionalized a set of relationships in which auditors have a motivation to please their clients, a state of affairs that effectively eliminates auditor independence. Moreover we have institutionalized these corrupt practices across the entire auditing industry. Under current law, auditing firms have financial incentives to avoid being fired and to be rehired by their clients. If an audit firm does not approve a client’s books, it runs a substantial risk of losing that firm as a client. Audit firms also profit greatly from selling nonaudit services, including business consulting services, to their auditing clients. In many cases, the profits that audit firms reap from selling nonaudit services far exceed their profits from audit services. In addition, as consultants they are in the compromised position of giving opinionated business advice to the same firms whose financial statements they are supposed to audit impartially. Finally, individual auditors often end up leaving the accounting firm to take jobs with client firms. In fact moving to work for a client is one of the most common job moves made by auditors. Together these conditions add up to a situation in which audit firms will be more profitable when they make their clients happy than when they honestly audit their books. As a result, “independent auditing firms” are not independent. All of these conditions that act against auditor independence were in place between Arthur Andersen and Enron. Again, consider this simple fact: Arthur Andersen was able to successfully maintain Enron as a client from 1986, soon after the energy company was founded, until the death of both organizations.
In my first publication on the topic of auditor independence, with Kimberly Morgan and George Loewenstein in 1997,4 we argued that decision makers typically view a conflict of interest as simply a choice between meeting one’s obligations and acting in a self-serving manner. A trader can either tell regulators about index fixing or request a fix in favor of her investments. The conflict is evident and the choice deliberate. This view of conflict of interest as exclusively intentional leads to the view that moral suasion or sanctions can prevent the destructive effects of conflict of interest. However, extensive research demonstrates that our desires influence the way we interpret information, even when we are trying to be objective and impartial. Most of us think we are better-than-average drivers, have smarter-than-average children, and choose stocks or investment funds that will outperform the market—even when there is clear evidence to the contrary. We discount facts that contradict the conclusions we want to reach, and we uncritically accept evidence that supports our positions. Unaware of our skewed information processing, we erroneously conclude that our judgments are free of bias.
Experiments going back fifty years have demonstrated the power of self-serving biases. In one famous study, Linda Babcock, George Loewenstein, Sam Issacharoff, and Colin Camerer had participants simulate a negotiation between lawyers for a plaintiff and a defendant.5 Pairs of participants were given the same police and medical reports, depositions, and other materials from a lawsuit involving a collision between a motorcycle and a car. The pairs were asked to try to negotiate a settlement to be paid by the defendant to the plaintiff. They were told that if they couldn’t reach a settlement, a judge would decide the amount of the award, and both parties would pay substantial penalties. Before negotiating, the participants were asked to predict the amount the judge would award the plaintiff if the negotiation failed. Participants were assured that the other party wouldn’t see his or her estimate and that the estimates would not influence the judge’s decision. Participants were also given incentives to be accurate. Nonetheless, on average, study participants representing the motorcyclist plaintiff predicted their side would receive awards about twice as large as those representing the defendant predicted.
My work with Don Moore and Lloyd Tanlu tested the strength of such conflicts of interest by giving study participants information about the potential sale of a fictional company. Study participants were asked to estimate the company’s value.6 They were assigned to one of four roles: buyer, seller, buyer’s auditor, or seller’s auditor. All participants read the same information, including information that could help them estimate the worth of the firm. Those acting as auditors provided estimated valuations of the company’s worth to their clients. Sellers submitted higher estimates of the company’s worth than did prospective buyers.7 More interestingly the auditors were strongly biased toward the interests of their clients: sellers’ auditors publicly concluded that the firm was worth far more than did buyers’ auditors.
Was this bias intentional, or were people engaged in unethical behavior without even knowing that they were doing anything wrong, what my colleagues and I call “bounded ethicality”?8 To figure this out, the auditors were asked to estimate the company’s true value, as assessed by impartial experts, and were told they would be rewarded for the accuracy of their private judgments. Auditors for the sellers reached estimates of the company’s value that, on average, were 30 percent higher than the estimates of auditors who served buyers. These data strongly suggest that the participants assimilated information about the target company in a biased way: being in the role of the auditor biased their estimates and limited their ability to notice the bias in their clients’ behavior. Simply being in a purely hypothetical relationship with a client distorted the judgment of those playing the role of auditor. Furthermore we replicated this study with actual auditors from one of the Final 4 large auditing firms. Undoubtedly a long-standing relationship involving millions of dollars in ongoing revenues would have an even stronger effect. Auditors are unable to distance themselves from their clients because of a bias in the direction of those who pay their bills.
When we first presented our empirical work on self-serving biases in the auditing process in the early part of the new millennium, the psychological research community generally responded, “We already know this, and we knew it a long time ago.” Psychological research had long shown that people who have a self-interest in seeing data in a particular direction are no longer capable of independence. In other words, we were accusing auditors of being human. Meanwhile members of the accounting field treated our results with disdain, since they assumed that auditors were fully independent. This group included heads of the major accounting firms, accountants in academia, and, as shown by their inaction, regulators. I believe this group could see bias only as an intentional process, and because they viewed auditors as honest, they assumed this made auditors immune from bias.
There have been many bad guys in the corporate world, people who intentionally engaged in illegal actions, including Bernard Madoff, Jeffrey Skilling, Kenneth Lay, and Andrew Fastow, to name just a few. But I honestly believe that far more harm has been done by the majority of us, who engage in bad behavior without recognizing that we are doing anything wrong and who watch others engage in unethical behavior and say nothing. Similarly the media found the stories of Marc Hauser’s and Diederik Stapel’s downfalls compelling, though only a tiny percentage of psychologists have been determined to have deliberately created fraudulent data. But even if such fraud is rare, that should give us only false comfort about the integrity of our data. In fact a more important story is emerging in academia: well-intentioned researchers have been undermining the integrity of their work and, indeed, the entire field without realizing that we were doing anything wrong.
In quantitative studies in the social sciences, it is common for researchers and peer-reviewed journals to use a very specific criterion to determine whether a result is “statistically significant.” The criterion is whether the p-value, a statistical value, is less than .05; this means that there would be no more than a 5 percent probability that a given result could have occurred by chance. Scientists use many different statistical tests, but the “p < .05” criterion remains dominant across almost all of these tests. Researchers understand that their results probably will need to meet the p < .05 standard to be published in leading scientific journals. But there are many ways that researchers can increase their chances of getting the p < .05 effect, namely by using what are known as “researcher degrees of freedom.”9
Imagine that a researcher has a hypothesis that men are more risk-seeking than women in their investments.10 The p < .05 criterion makes sense if the researcher defines one test of this hypothesis and decides in advance how many men and how many women will participate in the test. For example, you might bring men and women into the laboratory and ask them to make a set of investment decisions between stocks and bonds. Finding that men were more prone to choose stocks would be evidence in favor of your hypothesis. But what if you present your participants with stocks that have different levels of riskiness and also present them with bonds of different levels of risk? Now you are able to test whether:
1. Men are more likely to invest in stocks than in bonds as compared to women.
2. Men choose riskier stocks than women do.
3. Men choose riskier bonds than women do.
4. Men have a higher aggregated level of risk in their investments, as tested with three different aggregation methods developed by finance scholars (4a, 4b, and 4c).
Further, imagine that you run your experiment with fifteen men and fifteen women. You find that the results are in the direction you predicted, but not significant at the critical .05 level. You run the experiment again with fifteen other men and fifteen other women. Your results are now marginally significant (the p-value is between .10 and .05), so you run the experiment again with twenty men and twenty women. Finally, when you put the results of these three trials together, you find that men are significantly more likely than women to choose riskier stocks.
The basic idea behind this hypothetical example is that researchers can try lots of different outcomes to test the same idea; in research lingo, they can collect multiple dependent variables. They can collect a batch of data, and if the results approach significance, they can collect some more data, thus giving themselves multiple tries at getting p < .05. They can also decide to exclude certain data as outliers (i.e., strange responses that suggest participants didn’t understand the task) after collecting the data and seeing whether the exclusion would affect the results.
In 2011 research psychologists Joe Simmons, Leif Nelson, and Uri Simonsohn published a brilliant paper showing that using four such researcher degrees of freedom and adding a dose of creativity is extremely likely to turn up some confirming evidence at the p < .05 criterion level, even when the basic idea being tested is not true.11 Even using random data, when researchers test ideas in multiple forms, they have a chance that is far greater than 5 percent of finding the desired effect—and they can then publish the version that worked. Simmons and his colleagues also show that it requires very few researcher degrees of freedom to raise the likelihood to over 50 percent of getting a significant result, again even with random data. In short, their research shows that it is overwhelmingly possible to work within the field’s established protocols and arrive at desired but incorrect conclusions.
Peer-reviewed journals often support these questionable research practices by discouraging researchers from presenting their data and tests fully, ostensibly because this would take up too much journal space. And the greater the subjectivity of the field, the greater the potential for the use of questionable research practices. This suggests that the social sciences are particularly susceptible to such practices.
In a related paper, Leslie John, George Loewenstein, and Drazen Prelec conducted a survey of psychological researchers, using a complex procedure that induces people to respond honestly, and asked them about their use of a number of questionable research practices.12 These practices included (1) failing to report all of a study’s dependent measures (the outcomes being measured); (2) deciding whether to collect more data after testing whether the results were significant; (3) failing to report all of a study’s different conditions or versions; (4) stopping data collection earlier than planned because the desired result emerged; (5) rounding off a p-value in a favorable direction (for example, reporting that a p-value of .054 is less than .05); (6) selectively reporting studies that “worked” and not reporting those that didn’t; (7) deciding whether to exclude data after looking at the impact of doing so on the results; (8) reporting that an unexpected finding was actually predicted from the start; (9) falsely claiming that results are unaffected by demographic variables (such as gender); and (10) falsifying data. The last of these practices is data fraud of the type that has been connected to the Hauser and Stapel cases. Let’s focus instead on the nine other, arguably lesser, infractions.
The estimates obtained by John and her colleagues of the percentage of psychological researchers who engage in these questionable research practices varied between 36 and 74 percent for the first eight items and was 13 percent for item 9 and 9 percent for item 10. Even if these numbers are overstated by 100 percent, the use of questionable research practices is striking. The obvious conclusion is that there is a strong likelihood that many results reported in the research literature and then reported in the media are simply not true.
How did we get to this state of affairs in social science research? To start with, in recent years academia has become more competitive. Quite simply, today it is much harder to obtain a prestigious university position than it was thirty years ago, when I was a junior faculty member. The quantity of publications produced by the top new PhDs on the job market is astonishing. In addition, scholars at the top universities have the best chance of receiving media attention, large speaking fees, and rich book contracts, so the stakes of winning a coveted position are high. At the same time, journal editors want to publish “interesting” results. To publish more papers in a limited amount of space, many leading journals have shrunk the space devoted to details of the methods used in studies. Taken together, these factors conspire to motivate researchers to use too many degrees of freedom without noticing that they are doing anything wrong—at least until Simmons, John, and their colleagues pointed it out in their very visible publications. Many of us researchers have simply been following the advice of our mentors, the suggestions of journal editors, and the standard practices in the field, without stopping to think about how we were violating the very logic for the use of the p < .05 criterion. For many of us, these two excellent research papers were a call to change our ways and highlighted the need to clarify how research is conducted in our labs.
But not all social scientists share my enthusiasm for reforming research practices. In fact the authors of these two papers were attacked regarding minor details of their methodology. For example, John and her colleagues had only a 36 percent response rate to their survey, which led to questions about the representativeness of their sample. Thus it could be that their results were tragically overstated. But wait a minute: Who would be more likely to fill out a questionnaire about research ethics: ethical researchers or unethical ones? The answer to this question makes it quite obvious that the estimates of John and her colleagues are undoubtedly conservative.
John and her research team were also accused of acting unethically in order to get attention, and they were criticized for not understanding the basics of research methods. Both charges seem unlikely: the authors are prestigious scholars used to attention who have long lists of research publications evidencing a clear grasp of methods. The critics also focused on rare cases in which using a “questionable research practice” might be justified, accused the group of trying to destroy social psychology as a field, and tried to justify the suppression of the team’s findings as a need to protect the “in-group” (that is, to keep social psychology’s secrets from the press). Finally, critics argued that rather than taking steps to reform the field, more research was called for, specifically on the issue of researcher degrees of freedom. As I observed these defensive responses and delay tactics, I was horrified to note that they mirrored those that the U.S. tobacco industry, climate deniers, and audit firms have used to resist positive change in their industries.13
This issue of researcher degrees of freedom is not limited to particular social sciences or particular methods. Rather it offers a challenge for the field as a whole to determine how to conduct research with greater integrity. Put bluntly, we failed to manage our product line and have been criticized in the press for it. The problem of researcher degrees of freedom has been conflated with the problem of fraudulent data, which is a dramatically different issue. The confusion occurs because social scientists have turned a blind eye to their conflicts of interest. We may not have institutionalized the use of fraudulent data, but we did institutionalize many of the practices examined by John and her colleagues. This occurred as mentors taught doctoral students that many of these practices were standard, the research equivalent of driving five miles over the speed limit.
Consider which of these two options sounds more likely to succeed:
1. Auditors are prohibited from establishing durable, long-term cooperative partnerships with their clients, from providing nonaudit services to their clients, and from taking jobs with their clients.
2. After a variety of incentives that lead auditors to want to please their clients have been institutionalized, a complex set of legislative and professional incentives are put in place in an attempt to counteract the corrupting desire to please their clients.
This may seem like a silly choice: the former option obviously makes more sense than the latter. Yet for decades we have chosen the latter for political reasons.
While some auditors undoubtedly are aware that their behavior is corrupt, most auditors are much more frequently affected by self-serving biases that lead them, like all of us, to view data in a light that reflects what they want to see. Due to the often subjective nature of accounting and the close relationships that exist between accounting firms and their clients, even the most honest auditors can be unintentionally biased in ways that mask a client company’s true financial status, thereby misleading investors, regulators, and even the company’s management.
It is remarkably easy to identify the steps that would dramatically increase auditor independence:
1. Auditors should be hired under fixed contracts that stipulate true rotation of both individual auditors and the auditing firm. During the time period specified in the contract, the client should not be able to fire the audit firm. In addition, the client should not be allowed to rehire the auditor at the end of the contract (for a legally specified amount of time).
2. When a client changes auditors, personnel working on the audit for the outgoing auditing firm should not be allowed to move to the new auditing firm to resume work on the same client.
3. Auditing firms should not be allowed to provide any nonaudit services.
4. The auditing personnel for a particular client should be barred from being hired by the client for a specified period of time.
Opponents of audit reform will be quick to note, correctly, that these recommendations will create costs for audit firms and their clients. These opponents call for cost-benefit analyses before changes are made, knowing that conclusive, undisputed empirical evidence on auditor bias will be somewhere between extraordinarily difficult to impossible to find. My reaction is that the choice should not be between the status quo (which the auditing industry has invested many millions of dollars in lobbying efforts to create) and the reforms being proposed. Rather the choice should be between whether our society wants independent audits and whether we should eliminate the requirement to be audited. If we do want independent audits, it is time to recognize that, without a massive overhaul of the existing system, this goal will elude us. Society is currently paying enormous costs without getting the very service that the industry claims to provide: independent audits.
The auditing industry has spent tens of millions of dollars to block the creation of auditor independence in the United States. Only recently did accounting scholars become open to the obviousness of our work. To take one example, in 2011 the Management Accounting section of the American Accounting Association invited me to present my work on auditor bias as the keynote speaker at its annual meeting. This is quite a contrast to the prior fifteen years, during which the field of accounting largely denied the problem and helped perpetuate the financial crisis of 2008.
Similarly the recommendations for improving the integrity of psychological research and research in other social sciences is remarkably clear, and Simmons and his colleagues spelled out many of the needed changes.14 To name just a few of their recommendations, they advise researchers to decide how many observations they will gather before collecting data, to list all variables and conditions collected in their studies, and to be fully transparent about all of their decisions in published articles. But it is critical to note that, just as an auditing firm would lose market share if it were as tough as it should be, the researcher who voluntarily follows all of these rules is at a disadvantage in the publishing process, thus making this a problem at the industry level.
It is simply not reasonable to expect most junior researchers to voluntarily make the publishing process harder on themselves, even as they know that it remains easier for others. Instead journal editors and leaders of professional societies have a moral responsibility to act to change the rules of experimentation and reporting. We need to create a balanced, more honest playing field. Just as it was never reasonable to expect baseball players to reject steroids when they were obviously competing against steroid users (while Major League Baseball turned a blind eye), it is not reasonable to expect doctoral students and untenured professors to solve this problem. The burden should be on industry leaders to change the rules systemwide.
The problem of conflicts of interest is far from unique to audit firms and academia. Conflicts of interest can taint the recommendations of doctors, lawyers, investment advisers, real estate agents, and a host of other people on whom we rely for advice. What makes auditors fairly unique is that the only reason their profession exists is to provide independent assessments, and the conflicts of interest that infect the profession prevent that independence.
Note that I said fairly unique. Let’s consider the financial crisis of 2008. Fingers have rightfully pointed in many directions: at irresponsible banks, greedy home-buyers, speculators, the Democratic Congress for pushing to give low-income borrowers too much credit, and the Bush administration for poor decision making and regulatory neglect. All of these groups were affected by faulty judgment and conflicts of interest. But the group that most resembles the auditing industry is the credit-rating agencies.
Just as auditing firms exist to vouch for the financial condition of firms to outside stakeholders, credit-rating agencies exist to educate outside stakeholders of the creditworthiness of issuers of debt obligations. They are in the business of rating the riskiness of the investment securities they assess. For a financial firm to sell securities to the public, the securities must be rated by one of three major rating agencies: Standard & Poor’s, Moody’s, or Fitch Group.
As the housing bubble developed, debt issuers began to sell subprime and other high-risk home loans as mortgage-backed securities. Ample evidence documents that credit-rating agencies were too lenient in rating these securities and failed to be independent. It is now clear that the executives running the agencies gave AAA ratings (the highest level possible) to thousands of complex mortgage-related securities that they had reason to suspect were toxic. It is also clear that the credit-rating agencies lacked the independence that is the sole purpose of their existence. In the aftermath of the financial crisis, Representative Henry Waxman (D-CA), chairman of the House Oversight and Government Reform Committee, argued that “the story of the credit rating agencies is a story of colossal failure.”15
Standard & Poor’s, Moody’s, and Fitch are paid by the companies whose securities they rate. Like auditing firms, they are not accountable to the investors who have the most to lose from their leniency and independence failures. The Big 3 rating agencies made enormous profits by giving top ratings to securities and debt issuers. They were not rewarded for the accuracy of their assessments; in fact, for obvious reasons, the agencies with the most lax standards had the best chance of winning business from new clients and then faced a strong motivation to positively assess securities, creating a rapid race to the bottom. And just like the auditing firms, the credit-rating agencies were selling consulting services to the same firms whose securities they were rating.
This dysfunctional pattern was allowed to continue in the credit-rating agencies long after the financial collapse. In a September 17, 2013 article,16 the New York Times documented that Standard & Poor’s had made an executive-level management decision to lower their standards so that they could recruit more business. It is also clear that the market prefers lower standards: in response to Standard & Poor’s lowering its standards, its market share jumped from 18 percent soon after the financial crisis to 69 percent in 2013.17
If rating agencies have an incentive to please the companies they assess, unbiased assessments are not possible. Once again, like supporters of the auditing firms, defenders of the credit-rating agencies argued that the importance of ensuring a firm’s integrity would protect the agencies from issuing biased assessments. This belief in the integrity of the credit-rating agencies was, and continues to be, misplaced.
Lots of people don’t notice conflicts of interest. Professors, doctoral students, journal editors, and the leaders of the societies governing the social sciences didn’t notice that we were all engaging in questionable research practices in ways that collectively lowered the quality and integrity of the research being reported. Audit partners, auditors on the ground, clients, investors, and the Securities and Exchange Commission (SEC) didn’t notice that audits were not being conducted independently. The shareholders of the credit-rating firms, the employees that rate securities, the financial firms whose securities are rated, investors, and, again, the SEC didn’t notice that the credit-rating agencies weren’t performing their main function. Across these three examples, what is amazing is that entire industries were lulled into complacency.
When someone tells you, “This is just how it’s done in our field,” it should be a call to ask why it is done that way and whether there is a better way to do it. Too often we offer the commonality of an action in an industry as a reasonable explanation, even when it is clear that what is accepted practice is not necessarily correct or appropriate. The fact that the behavior is common institutionalizes it as normal; it doesn’t certify the behavior as the right thing to do. And when we fail to oppose such unethical behavior we become part of the problem.