BP’s Failure to Debias (Page 3 of 6)
Groupthink is a form of collective confirmation bias, and reflects inadequate airing of the pros and cons of competing alternatives, often because the group leader discourages devil’s advocacy. It is in this sense that groupthink operated on Deepwater Horizon the day of the explosion. However, as we now argue, as the day progressed confirmation bias was particularly pronounced.
For the next few hours, Transocean workers conducted two pressure tests, a positive pressure test and a negative pressure test. The positive pressure test involves increasing the pressure inside the well by pumping fluid, to see whether fluids leak from the well. This test was begun at about noon, and the result was favorable.
The negative pressure test involves decreasing pressure from the well, to see whether fluids leak out. The Transocean crew proceeded to remove mud from the well in order to decrease pressure, and by 5 p.m. had commenced the test itself. It was at this stage that confirmation bias loomed large. The test results were unusual, and Transocean workers struggled to interpret the readings. Pressure built up unexpectedly with no clear reason as to why. Despite his earlier resistance, Harrell judged the issue to be non-problematic. He had a valve at the top of the blowout preventer tightened, which seemed to address the issue. However, other Transocean workers were not persuaded that the problems had been resolved. For example, Wyman Wheeler, who supervised the drilling crew for twelve hours per day, was not convinced that all was in order. Yet, when Wheeler’s shift ended at 6 p.m. his replacement, Jason Anderson, assured both his Transocean co-workers (and for that matter his BP colleagues) that the pressure readings were normal. Anderson suggested an alternative hypothesis called “U-tubing” (or “bladder effect”) for the observed readings.16
BP managers also disagreed with each other. Donald Vidrine was the BP manager due to relieve Kaluza at 6 p.m. Despite having made the argument for removing so much mud, Kaluza was uncomfortable with the results of the test, and Vidrine was especially concerned about a surge of gas. For that reason, Vidrine decided to order a second test, somewhat different from the first. The results of the second test were especially perplexing. Gauges on the main pipe indicated nonzero pressure, which signaled a problem, although a smaller tube leading up from the well showed no pressure, a sign that the well was stable. Notably, the two pipes were connected and should have featured the same pressure. Vidrine consulted with a BP superior, Mark Hafle in Houston, who assured Vidrine that had there been a kick in the well, it would have already been detected.
The National Commission report made the following comments about the judgments just described, stating “The Well Site Leaders and crew never appear to have reconciled the two different pressure readings. The “bladder effect” may have been proposed as an explanation for the anomaly—but based on available information, the 1,400 psi reading on the drill pipe could only have been caused by a leak into the well. Nevertheless, at 8 p.m., BP Well Site Leaders, in consultation with the crew, made a key error and mistakenly concluded the second negative test procedure had confirmed the well’s integrity. They declared the test a success and moved on to the next step in temporary abandonment… (p. 108) It appears instead they started from the assumption that the well could not be flowing, and kept running tests and coming up with various explanations until they had convinced themselves their assumption was correct.” (p. 119)
The decision backdrop pertaining to the negative pressure test, what behaviorists describe as base rate information, featured a well with a troubled history. In an email message one BP manager described Macondo as a “nightmare well.” At various times, the drill got stuck. At other times, the well “kicked,” meaning gas shot back through the mud, sometimes at an alarming rate. A Transocean employee interviewed on CNN stated, “There was always like an ominous feeling. This well did not want to be drilled. ... It just seemed like we were messing with Mother Nature.”
At approximately 7:50 p.m. Vidrine instructed that a call be placed to BP engineers in Houston stating he was satisfied with the results of the negative pressure test.17 His decision was taken against the backdrop of negative base rate information: Hafle’s judgment from Houston about there being no gas surge, Anderson’s competing U-tubing hypothesis, a crew anxious to move to the next project, and the pressure exerted by dealing with the project being late and over budget. Interestingly, in his 30 years of experience on rigs, Vidrine had never seen a case of U-tubing; he had only heard about it. Our sense is that confirmation bias and aversion to a sure loss exerted strong influences18 An alternative view is that the explosion was simply a tail event, and that our sense reflects hindsight bias.
In the two hours between Vidrine’s message to Houston and the first explosion, unfavorable signals continued to be generated. For example, electronic data reviewed by investigators after the explosion showed that drill pipe pressure was increasing, a significant anomaly signaling the strong possibility of a gas kick. The National Commission suggests that the Transocean crew might have missed the signals because they had become distracted by other tasks.
According to Shell engineers Leimkuhler and Hollowell, the April 20 test would have been a point of high risk in the process, as the mud restraining any escaping gas and oil would have been removed, thereby providing a potential escape channel for that gas to make its way to the ocean surface.
Again, the base rate for Macondo was that it was a difficult well. The well design featured fewer barriers than the design used by competitors Shell and Exxon. The project was behind schedule and over budget, and a modified procedure for a critical test was introduced at the last minute. Was April 20 a time for focusing resources on the tasks at hand, or was it a time to introduce distractions?
BP chose the distracting route. Ronald Sepulvado, the BP manager in charge of the rig was on shore that day for a training program with his phone switched off, Transocean’s Harrell, and his second-in-command Randy Ezell, had spent much of that day hosting executives visiting the rig. The visiting executives included BP’s Pat O’Bryan, who had recently been appointed vice president for drilling in the Gulf of Mexico. Ironically, the agenda included commending the crew for its safety record and to discuss coming maintenance. Also ironically, O’Bryan was an expert in detecting gas leaks in oil wells.
4.3 Excessive Optimism and Overconfidence in BP’s Crisis Management
We now come to issues that received the most media attention, and with which people are most familiar. After the explosion of Deepwater Horizon, and the failure of the BOP to prevent oil and gas from escaping from the well, there were serious concerns about the environmental impact on the Gulf, especially on the fishing industry and on recreational activities on US beaches. BP’s CEO at the time was Tony Hayward. He was under pressure from the US government and the world media. Did his behavior exhibit psychological pitfalls? To answer this question, consider a series of statements Hayward made in May.
On May 14, Hayward told the British newspaper the Guardian, “The Gulf of Mexico is a very big ocean. The amount of volume of oil and dispersant we are putting into it is tiny in relation to the total water volume.” On May 17, BP inserted a siphon into the ruined riser pipe and began to collect 1,000 barrels of oil per day. This led Hayward to say, “I do feel that we have, for the first time, turned the corner in this challenge.” That siphoning effort was later abandoned. On May 18, he told the BBC, “I think the environmental impact of this disaster is likely to have been very, very modest.” In our view, these comments all reflect excessive optimism and overconfidence about ability.
Although Hayward’s May 14 statement is technically true, it is seriously misleading in so far as impact is concerned. Here is what is technically true. According to government estimates, between April 20 and July 15 when BP placed a temporary cap on the well, approximately 4.9 million barrels (206 million gallons) spilled into the Gulf from Macondo. In contrast, the volume of water in the Gulf of Mexico is approximately 643 quadrillion (6.43 x 1017) gallons. In addition, government estimates indicated that 74 percent of the oil which leaked subsequently evaporated, broke up, or was skimmed or burned off.
However, the ratio of the spill to the volume of water in the Gulf is misleading as a measure of the damage caused by the spill. BP eventually estimated that the cost of the spill would be US$32 billion and set aside US$20 billion in reserves.19 As oil from Macondo washed up on beaches in Louisiana, Mississippi, Alabama and Florida, the spill severely impacted the Gulf Coast economy, and threatened its ecology. Federal and state authorities shut down Gulf fisheries. The Federal government instituted a temporary federal ban on deepwater drilling, thereby idling oil workers. Scientists warned that the Gulf wetlands, which are pivotal in its ecology, were at high risk. In addition, as of September, the amount of spilled oil which has not disappeared remains controversial. A team of researchers from the University of Georgia announced that they had identified a two-inch thick oily layer coating the ocean floor at locations stretching up to 80 miles from the Macondo wellhead, which they believe stems from the BP spill.
Throughout much of the cleanup effort, statements from BP executives reflected excessive optimism and overconfidence. For example, on June 8, BP Chief Operating Officer Doug Suttles stated that the spill “should be down to a relative trickle by Monday or Tuesday” (Sappenfield, 2010). In a major effort at the end of May, called “Top Kill,” BP sought to plug (“kill”) the well from the ocean floor (the “top”). “Before ’top kill’ started, the company’s executives were genuinely optimistic that it might work.” Hayward said that “top kill” had a 60–70 percent chance of stopping the oil flow. Top kill failed (Crooks, 2010).
Excessive optimism and overconfidence were persistent features of BP’s public announcements. In the first weeks following the Deepwater Horizon explosion, BP estimated a spill of 1,000 barrels of oil a day. Soon after, they raised their forecast to 5,000 barrels daily. In the second week of June, independent experts suggested that a more precise estimate could be between 35,000 and 60,000 barrels a day.
The excessive optimism of BP’s management was also related to the real dimension of what they were facing in the Gulf. They deeply underestimated the size of the oil spill flow rate from the well, and then did not try to adjust it.20
As for overconfidence, in seeking permits to drill in the Gulf, BP claimed it could handle a leak of 250,000 barrels of oil per day. “Those claims were later shown to be ludicrously overconfident” (Crooks, 2010).
5. MISSED SIGNALS
A 4x4 analysis of BP’s corporate culture, conducted after events at Texas City and Alaska, but before the explosion on Deepwater Horizon, points to excessive cost cutting, weak risk management practices, and high risk exposure. These traits were manifest in BP’s decisions about the Macondo well. Many analysts, investors, regulators, and BP’s own executives missed the signals. This section discusses the missed signals before April 20 and the conclusions reached thereafter.
5.1 Hayward’s Tenure: Cost Cutting, Safety, and Culture
BP’s value destructive excessive cost cutting and excessive risk taking were traits that deepened, if not emerged, under the leadership of John Browne who was CEO from 1995 to 2007. As the discussion in section 2 pointed out, BP management not only took risks with safety by neglecting aging equipment, but pressured or harassed employees to refrain from reporting problems, and to cut short or delay inspections in order to reduce production costs. In this regard, the report on the Texas City disaster led by former US Secretary of State James Baker, stated, “BP has not provided effective process safety leadership and has not adequately established process safety as a core value.”
In the wake of accidents at its operations in Texas City and Alaska, Tony Hayward replaced Browne as CEO, with the charge of improving safety at BP. When he was appointed in 2007, Hayward sought to reduce the complexity of the company. He restructured divisions and cut administration and support functions.21 In this regard, he set up a new risk management system to standardize safety practices and to prevent other accidents from occurring.
Hayward appears to have believed that safety at BP was trending upward, and that the explosion on Deepwater Horizon was effectively an outlier. In a memorandum to employees just after the explosion, Hayward stated, “This accident has been a terrible exception to that trend and we must learn the lessons from it. But at the same time, it does not invalidate all the hard work you have put in to improve our safety standards around the world. Safety is our first priority. It will remain so” (Lustgarten, 2010).
Although safety might have been first priority at BP in Hayward’s mind, the record shows that as of June 2010, BP had 760 OSHA fines for “egregious, wilful” safety violations. By way of contrast, Exxon Mobil had just one violation (Sverjensky, 2010). Anecdotally, Lustgarten (2010) describes an incident in 2008, a time during Hayward’s tenure as CEO, in which BP failed to deliver final “as built” design drawings to crews operating deepwater rigs in the Gulf. These drawings are considered an essential safety component because they not only provide the basis for establishing that equipment operates properly, but also serve as instruction manuals in case of emergencies. Lustgarten point outs that an independent contractor met with resistance when he raised the issue of the drawings with BP engineers and management, who he suggests were seeking ways to reduce costs by several million dollars. The contractor subsequently lost his contract.
In May, Congressional hearings into the explosion of Deepwater Horizon led Bart Stupak, chair of the oversight and investigation committee to point out that BP’s corporate culture was characterized by excessive cost cutting and excessive risk taking. He stated, “I am concerned that the corporate culture from BP CEO Tony Hayward down to Chairman and President of BP America Lamar McKay, and Chief Operating Officer Doug Suttles and possibly down to the leadership on exploration rigs, reflects a willingness to cut costs and take greater risks.”
Stupak’s statement provides support for Shefrin’s 2008 assessment of BP. Further support for this characterization of BP comes from personnel who worked on Deepwater Horizon. Those personnel stated that BP repeatedly cut corners22 and persevered despite warnings about safety. One worker pointed to a dichotomy in respect to safety. He stated that one day he was scolded for standing on a bucket on the rig. Yet the next day, a crane violated safety policies by operating in the face of high winds (Bronstein and Drash, 2010).
Accounts of the sort just described led to an interesting observation by David Michaels, assistant secretary of labor for occupational safety and health. He stated, “The way safety is measured is generally around worker injuries and days away from work, and that measure of safety is irrelevant when you are looking at the likelihood that a facility like an oil refinery could explode. This is comparable to saying that an airline is safe because the pilots and mechanics haven’t been injured.”
5.2 Analysts: Availability Bias and Confirmation Bias
Analyst coverage of BP is illuminating. We examined reports between October 2006 and September 2010 to assess analysts’ perceptions and recommendations, with special emphasis on 130 reports from 27 brokerage firms during the period August, 4, 2009 to September, 17, 2010. See Table 1 for a summary of reports issued between April 20 and September 10, 2010. After reviewing the reports, we move onto our general conclusion that analysts lacked a framework for assessing risk management practices and corporate culture at BP, leading them to issue excessively optimistic recommendations reflecting availability bias and confirmation bias.
One of the strongest results in the literature on financial analysts is that analysts’ recommendations tend to be biased upward. This feature has been often been explained by potential conflict of interests faced by analysts. In this regard, analysts working for financial intermediaries with actual or potential business relationships with the companies being covered have an incentive to issue positive recommendations in order to encourage business dealings between their employer and the covered companies. Positive recommendations issued by analysts may also attract new businesses for their employer. Evidence shows that stocks positively recommended by affiliated analysts tend to perform worst than the ones recommended by independent analysts (Michaely and Womack, 1999; Barber, Lehavy and Trueman, 2007). As a general matter, independent analysts’ recommendations display less upward bias (Malmendier and Shanthikumar, 2007).
16. “U-tubing” refers to cases where the downward pressure from mud (heavy drilling fluid) located between the drill pipe and the well walls surrounding it pushes seawater back up the drill pipe.
17. Rep. Henry Waxman said the oil company told the Energy and Commerce Subcommittee on Oversight privately that the well failed the key pressure test. “Yet it appears the companies did not suspend operations, and now 11 workers are dead and the Gulf faces an environmental catastrophe,” Waxman said, asking why work was not stopped on the well. hypothesis, a crew anxious to move to the next project, and the pressure exerted by dealing with the project being late and over budget. Interestingly, in his 30 years of experience on rigs, Vidrine had never seen a case of U-tubing; he had only heard about it. Our sense is that confirmation bias and aversion to a sure loss exerted strong influences.hypothesis, a crew anxious to move to the next project, and the pressure exerted by dealing with the project being late and over budget. Interestingly, in his 30 years of experience on rigs, Vidrine had never seen a case of U-tubing; he had only heard about it. Our sense is that confirmation bias and aversion to a sure loss exerted strong influences.hypothesis, a crew anxious to move to the next project, and the pressure exerted by dealing with the project being late and over budget. Interestingly, in his 30 years of experience on rigs, Vidrine had never seen a case of U-tubing; he had only heard about it. Our sense is that confirmation bias and aversion to a sure loss exerted strong influences.
18. Sadly, Jason Anderson was one of the eleven who perished during the accident.
19. Analyst reports suggest these amounts might overstate the value of the damage.
20. In doing so, they displayed “anchoring bias” that leads people to remain mentally anchored to a specific reference point, and not adjust sufficiently. Anchoring is related to conservatism and it induces poor planning, and thus to insufficient response in case of problems. BP’s executives remained anchored to their initial estimates, and did not want to adjust them subsequently, underestimating the real size of the problem.
21. Under Browne, BP made significant investments in renewable energy. However, Hayward reduced those investments.
22. As a matter of fact, in February 2009, Hayward told reporters, “The mantra in BP today is ’Every dollar counts’” (Crooks, 2010). In 2009, BP implemented a US$4bn cost reduction. While production increased by only 4%, the unit production costs reduced by 12% (BP, 2009, p. 84).