WorldCom CEO Bernard Ebbers

The paper came out a few months ago, but NPR’s recent story reminded me of it.  Essentially, a professor and graduate student at Stanford’s Graduate School of Business tried to identify the linguistic signaling used by CEO’s and CFO’s during earnings calls that could be used to reliably predict when their companies were cooking the books (PDF of full report here).  Their methodology was to analyze thousands of transcripts from earnings calls and then look for relationships between words and phrases and those companies that later had to restate their earnings.

Here’s a summary of their findings:

Using conservative statistical tests, we nd that the out-of-sample performance of the models that are based on CEO or CFO narratives is signicantly better than random by 4%- 6% (with 50% – 65% accuracy) and provides a signicant improvement to a model based on discretionary accruals and traditional controls. We find that answers of deceptive executives have more references to general knowledge, fewer non-extreme positive emotions, and fewer references to shareholders value and value creation. In addition, deceptive CEOs use signicantly fewer self-references, more third person plural and impersonal pronouns, more extreme positive emotions, fewer extreme negative emotions, and fewer certainty and hesitation words.

CEOs overuse words like “we” and “our team” versus “I”, since lack of “I” indicates that the speaker doesn’t psychologically own their statements or feel “responsible for what [they] are saying”.  This implies that investors and officials should regard such skewed ratios as possible indications of fraud.

Additionally, CEO’s overuse of superlatives to describe the current state and future outlook of their companies was also found to be diagnostic (from NPR):

Lying CEOs also tend to use a lot of words that express positive emotion — things are fabulous and fantastic and extraordinary.

Here’s what Enron CEO Kenneth Lay said when he addressed his employees at a time when the company was about to implode: “I think our core businesses are extremely strong. We have a very strong competitive advantage. Of course, we are transferring this very successful business model and approach to a lot of new, very large markets globally.”

Words like that can be a form of overcompensation.

“If all my speech is ‘fantastic,’ ‘superb,’ ‘outstanding,’ ‘excellent’ and all my speech sounds like a big hype — it probably is,” Larcker says.

It is interesting to think about language as a screener and/or signal.  One can think of a host of applications outside of earnings calls (politics, criminal justice, job interviews, etc).  Essentially, the language used by executives might be taken as a costly signal by investors.  Rather than cheap talk, the structure and nature of the words themselves may serve as a credible signal.  This makes intuitive sense, since only CEOs that are truthful in their representation of the company are likely able to provide detailed, certain answers that provide the good, bad, and ugly of a company’s performance.  However, while this signal may allow honest CEOs to communicate their veracity it will likely lead to a lot of false positives (i.e. suspected fraud), since the type of language cited by the researchers could be deployed for reasons other than fraud.

The research team built on the physical signal of the color red, which is displayed on the body of female chimps prior to ovulation:

Male chimps [responded to this signal] by masturbating and attempting to mount them. [...] Daniela Kayser‘s team found that when a lady wears red it prompts men to ask her more intimate questions and to sit closer to her.

The key finding was that men who’d viewed the woman wearing red opted to ask more intimate questions.

Kayser and her colleagues said their findings are consistent with evolutionary accounts of human attraction and have obvious practical implications. ‘It appears that women would do well to wear a red shirt or dress when preparing for a date with a desirable man, and women may be particularly successful in online dating when they post a picture of themselves in red apparel. More generally, our findings should be of considerable interest to fashion consultants and product designers, as well as marketers and advertisers.’

The last part of that quote is interesting. With chimps, the red display isn’t simply a signal (an action that the sender chooses to send and can potentially manipulate), but a costly signal or screening device (an action that is not under the control of the sender or one that is so costly only a truthful sender could deploy it). The chimps can’t turn the display on and off to their advantage.  Women, on the other hand, can choose to wear red whenever they like.  If the finding becomes well known–that women who seek to be viewed as more desirable will wear red in an attempt to subconsciously influence men–the signal may likely lose it’s effectiveness and be discounted by men.  Unlike chimps, men can use reason to temper the effect of their emotions (well, sometimes).

Also, as the British Psychological Society notes, you run the risk of some pretty humorous situations–like entering bars that are awash in red-clad patrons:

Were these recommendations to be heeded widely, it raises the comical prospect of city bars and night-clubs being filled entirely with red-clad women and men, like rival sports teams arriving for a match only to discover they’re both wearing the same strip. Yes, the men in red too, because another recent study by the same research team found that men wearing red were rated as more attractive and high-status by women.

(Via Bob Sutton)

If we come to her door in our pajamas–or at least different clothes (e.g. sweatpants, etc) than when she last saw us–she takes it as a signal that we are in for the night.  If we were going out or not going to bed soon we would still be in our regular clothes that we wore earlier.  If we haven’t changed, she probes–”why aren’t you in your jammies?”  This let’s us know that she suspects we aren’t in for the night.  It also means that she will likely spend a fair amount of time looking out her window to see if our cars stay in the driveway before she will settle in and go to sleep.  Now, putting on pajamas isn’t that costly of signal–there is nothing stopping us from putting them on and then changing back into regular clothes to leave the house or host guests.  (However, in all honestly this isn’t likely to happen.)

The lesson here is that a) the idea of seeking out signals is intuitive for people and we start at a very early age, and b) rather than fight with our daughter about going to bed we might be better served just changing into our pajamas out the outset to demonstrate to her that we aren’t leaving the house, no one is coming over, and we are also getting ready for bed.  She may not believe our words, but she seems to believe the signal that she’s identified.  Leveraging that signal can lead to better communication and the outcome that we want.

Allegedly, the anxious folks at these various luxury houses are all aggressively gifting our gal Snookums with free bags. No surprise, right? But here’s the shocker: They are not sending her their own bags. They are sending her each other’s bags! Competitors‘ bags!

Call it what you will — “preemptive product placement”? “unbranding”? — either way, it’s brilliant, and it makes total sense. As much as one might adore Miss Snickerdoodle, her ability to inspire dress-alikes among her fans is questionable. The bottom line? Nobody in fashion wants to co-brand with Snooki.

Read the full story here.  Hat tip Marginal Revolution.  This is awesome on so many levels.

Essentially, luxury brand marketers are engaging in a sort of counter-signaling campaign.  Rather than signaling the exclusivity and virtue of their brand through the conspicuous consumption of their products by admirable celebrities, these operatives are tying their competitors’ products to an undesirable celebrity in an attempt to downgrade the status of those products.  Genius.

In the clip below (first 30 seconds), two men are “tearing money”–literally taking cold hard cash and tearing it to pieces.  Each time one of them tears some cash the other ups the ante and tears even more money:

It’s a pretty creative way of sending a costly signal–demonstrating one’s type through a costly act that only a strong actor could afford to send.  In this case, the two men are engaging in a kind of relational signaling–the one that can tear the most money will prove that they are stronger of the two.

There is a stream of research on negotiation that shows the strategic expression of anger is effective, apparently because it is taken as a sign you are “tough” and thus leads your intimidated opponent to make concessions.  A new study by Hajo Adam and his colleagues suggests that this may be a culturally specific finding, which applies to people of Western descent but not necessarily others.  In a pair of studies that compared people of European descent to people of East Asian descent they found, in both a hypothetical and a more realistic negotiation, that people in the two groups had opposite reactions to negotiating with an angry opponent:

“Western-ancestry students were more likely to make a concession to their negotiation partner whereas the East-Asian ancestry students were less likely to do so.”

Various iterations of the study pinpointed that the expectation of what was considered “culturally appropriate” was the key determinant of subject reaction.

Intuitively, I think this makes a great deal of sense.  Too often, we let rational models of negotiation and interaction dominate our thinking on the subject and forget that context, culture, and psychology have a substantial impact on how such strategies will work in practice.

The strategy of expressing anger and displaying strength and toughness can be traced back to the earliest work of Schelling and indirect coercion.  Actor A wants Actor B to change the status quo (i.e. make a concession).  To do so, Actor A sends a signal that conveys their toughness and resolve which should influence Actor B’s behavior given this new information about A’s type and what they are capable of.

From a rational-actor perspective, this makes sense.  However, problems can arise when we try to implement such a strategy in practice.  People don’t always conform to the rational-actor model.  In the case of the research by Adam, et al., the audience’s reaction is influenced by their perception of what is appropriate versus what is utility maximizing.  (One might say that what is considered utility-maximizing is culturally determined.)  This means that when you send a signal you have to take into account the perspective of your audience (culture, psychology, etc).  It may not always be the determining factor, but it may significantly affect how your signal is received and the behavior it provokes.

However, I think there are limitations to the generalizability of the research since it didn’t address a crucial element: multiple audiences.

Signals are rarely sent in a vacuum.  Particularly when it comes to business, people rarely interact in a purely one-on-one setting, where their actions are guaranteed to stay hidden from any third parties.  Typically, the signals we send will be seen and interpreted by multiple audiences and this can significantly complicate negotiations.  Public bullying may backfire and cause the target to dig in their heels instead of offering concessions.  Signals must not only be clear enough to minimize the variance around their interpretation.  They must also avoid creating a situation that actually makes it harder for the target to behave in the way the sender wants them to.

The research by Adam, et al., looked at pairs of negotiators separated from other observers.  If they had introduced third parties my guess is that their results would have been different.  Backing down to an aggressive threat would likely be viewed as unacceptable across both Western and East Asian subjects for related reasons.  For many Westerners, backing down to a threat would signal weakness and a lack of resolve.  If third parties are present and see this they may assume that the subject that made the concession could be bullied in a similar manner in the future.  In order to avoid being challenged in this way by multiple actors down the road, the subject is more likely to stand firm and avoid giving any concessions under duress.*  A similar dynamic is likely to play out with those of East Asian decent given the importance of “face” and honor in those cultures.  When additional subjects are introduced into the experiment the results are less likely to break down among cultural lines.

(Thanks to Andrew Chen for pointing to Sutton’s post)

*Some have dubbed this the “Culture of Honor” and theorized that, in the United States, it explains why Southerners may be more prone to violence than Northerners.  This culture is structurally determined, however.  Cultures of Honor are likely to develop where a) individuals are at economic risk from other citizens (e.g. where what is most valuable is easy to transport), and b) where the state is weak or nonexistent and cannot reliably protect personal property.  The argument goes that, for structural reasons, herding societies are the most prone to develop a Culture of Honor and herding (as well as settlers immigrating from herding cultures) was more prevalent in the South.  Therefore, a Culture of Honor was inculcated amongst Southerners but not Northerners.

The problem with behavioral signaling is that actors have incentives to pass themselves off as something they are not (e.g. capable, smart, committed, etc).  This has fueled research in economics and political science, with researchers searching for situations and mechanisms that make signals reliable–i.e. the signal can only be sent by actors that have certain characteristics, and therefore are separated from actors that do not.

But what about physiological signals?  If behavioral signals are subject to manipulation, why not focus on the physiological signals sent by the human body?  For example, if we want to know if an actor plans on following through with a commitment it would be helpful if there were physiological signals given off by the body (e.g. blood pressure, rapid eye movements, body heat, facial movements) that were highly correlated with the act of lying.  Dr. Paul Ekman has been conducting breakthrough research in one of these areas for decades, focusing on what he terms “micro-expressions”–physical manifestations of emotions that “leak out” when someone tries to conceal them.  (Ekman’s work is the inspiration behind the FOX television series Lie to Me.)

Physiological signals would seem to be more reliable since faking these types of signals is conceivably very difficult, if not impossible.  However, there is another concern one mus take into account regarding physiological signals, and their practical application in areas such as transportation security: false positives.  While physiological signals might make it difficult for someone (say, an actor intent on hijacking a plane) to hide their intentions, the same physiological signals that indicate ill intent could also be present if someone is generally nervous about flying, being questioned by authorities, etc.  The key to using a signal as a diagnostic tool is it’s ability to reliably separate types of actors.  If the same signal can be intentionally sent or passively expressed by actors of different types (i.e. hijacker versus nervous flyer), then the signal loses its power.

Fast Company recently profiled some new technology from WeCU Technologies that proposes to reliably diagnose risky airline passengers:

CEO Ehud Givon compares the function of WeCU’s detection system with a doctor’s diagnosis. The technology is relatively simple: a synthesis of electronic sensors and knowledge gleaned from behavioral studies. During a routine act, such as check-in at an airport kiosk, travelers will be subjected to a near-invisible stimulus that will trigger physiological responses among those who are concealing something. Sensors hidden in the kiosk will pick up the cues and alert security officers. WeCU’s boldest claim is that its system can weed out the mal-intentioned from the merely stressed out.

WeCU claims that they’ve developed their measures in such a way as to avoid the problematic false positives that have long provided a barrier to the deployment of these types of systems.  It will be very interesting to see how they fair once introduced into a live environment.  If the system is able to minimize the number of false positives as well as the number of false negatives (the number of terrorists that evade the system) it will be a big step forward in the practical application of physiological signaling.

as part of having a phone with a keypad (not even keyboard) I went back to T9 (you remember it, predictive type for SMS messages).  The struggle to send a simple message made me think about how we are likely going to need to start add elements to our interfaces that actually make tasks harder, not easier. Imagine how many fewer emails you’d write on your iPhone if you were forced to use T9 for instance.

While it’s not the perfect parallel, something like WriteRoom lets you block out everything else when you’re trying to write (Pages now has a full screen option as well, actually) and I remember hearing about an application from a few years ago that turned off your WiFi until you restarted the computer so you could get something done. It’s funny to think that we’ve reached a point where things are so easy that we need to start making them hard again.

There are lots of these ‘nudging’ services available (e.g. stickK and LeechBlock are just two that come to mind) where people can structure the incentives around an activity to make it’s completion more likely.

This reminds of a much older concept related to relaying the credibility of one’s commitments: tying one’s hands.  The premise, articulated famously by Nobel Prize winner Thomas Schelling in regards to nuclear deterrance, is that in order to convince another of one’s resolve to carry through on an undesireable act one must design mechanisms that would make it impossible to avoid following through.  Schelling cites an episode from the Cold War to illustrate how this (intentional or unintentional) dynamic works:

This is the postion Chiang Kai-shek got himself into, and us, when he moved a large portion of his best troops Qeumoy.  Evacuation under fire would be exceedingly difficult; if attacked, his troops had no choice but to fight, and we probably had no choice but to assist them.

Schelling as often spoke about the utility of tying one’s own hands on purpose to prevent bad behavior or force good behavior.  (On many occasions, Schelling has spoken about using such strategies to quite smoking,)  Like Cortez, burning the boats provides a fool-proof mechanism for ensuring full committment–press on with a difficult task since there is now no alternative.

I don’t know if we need these services more than ever because of technology.  Rather, I think technology makes it easier for people to employ such strategies on a daily basis and use them to their advantage.  We can’t all burn the boats like Cortez, but technology gives us a reasonable approximation.

Traditionally, credit rating agencies–such as S&P, Moody’s and Fitch–served as a sort of signal or barometer for investors as to the level of risk attached to specific corporate bonds.  Over time, they began rating additional types of investment products, including new financial instruments such as collateralized debt obligations (CDOs) that consisted of massive bundles of subprime mortgages.  Amazingly, these agencies rated these products AAA–the highest possible rating, indicating “the highest level of capacity of the obligor to honor its financial commitment on the obligation.”  The importance of this cannot be understated.  What makes this particularly troublesome is that the firms that issued the debt were the ones paying the CRA’s for the ratings, creating a significant conflict of interest.  Common practice, sure.  But given what transpired in this particular case such a conflict cannot help but be scrutinized.

What was once a reliable indicator for measuring the level of risk attached to an investment has now come into question as possibly nothing more than a rigged instrument.  Now, US Congressional leaders are focusing aspects of their financial reform bill on the CRA’s.  Senator Dodd’s version of the bill included a number of provisions aimed at restoring trust and confidence in these vital agencies.  Just yesterday, the Senate passed an amendment offered by Sen. Franken (MN) that would require a newly-formed government body to assign the task of rating a specific asset-backed security to a specific CRA.  The hope is that such a move would end the practice of ratings shopping that many believe led to creation of perverse incentives by CRA’s and the misleading rating of unstable securities.

I am not by any means an expert on the CRA industry (I know just enough to be dangerous, I guess), but here is my take.

The overall goal should be to create a ratings system that is both widely used and viewed as credible by the investment community. The major issue here is that we need to end the conflict of interest (or perception thereof) that exists when you have a ratings agency being paid to evaluate and rate the level of risk of a complex financial instrument by the very organization that is issue the instrument.  I think the Franken amendment is coming from the right place, but in the end it might be misguided.  Creating a federal agency to determine who gets to rate which security may continue and further perpetuate the false sense of security that investors have in the accuracy of these ratings.  The governmental seal of approval can also induce perverse behvavior.

What I would suggest is the creation of an agency that would manage the oversight of CRA’s in a way that still allows for the market to not only evaluate risk but also evaluate the CRA’s and the ratings they issue.  Here are a few approaches that I think would be helpful:

• (Good) Require multiple ratings for any single security: As with anything, there is power in numbers.  If the government agency is going to determine who rates a single security why not involve more than one firm?  Why not three CRA’s and the eventual rating is some kind of weighted average of all three?  This could help remove the false sense of security from the government seal as well as the conflict of interest between issuer and CRA.
• (Better) Provide greater transparency into the ratings process: A better understanding of how the CRA came to it’s conclusion would allow investors to contextualize the rating and how reliable that rating is.  How we cast the die is important when it comes to risk analysis.  Knowing the model, the assumptions, and the data sources provides investors a chance to weight how reliable the rating actually is.  Even if the issuer pays off the CRA investors can independently evaluate the reliability of the particular rating.  Requiring a standard period of review before the security is available for sale would allow for independent evaluation.
• (Best) Institute rigorous evaluations and independent rankings of ratings agencies (Moneyball-approach): Right now I believe the federal government can register a CRA, but lacks the tools and ability to continuously evaluate and possibly deregister a poorly performing CRA.  What we should be able to do is come up with a way to rank CRA’s against each other based on the reliability of their risk assessment for securities over time.  Think of it as a Value-Over-Replacement-Rater or VORR, similar to the way that professional athletes are now rated and their value determined (Value-Over-Replacement-Player or VORP).  Given the mass of historical data on securities ratings and their performance it shouldn’t be too hard to create a VORP-like rating system that could be updated in real-time as the value of securities rise and fall.  This way, it isn’t simply a government’s stamp of approval or a firm’s reputation that determines the level of trust investors should have in a CRA’s ratings.  Instead, investors can see in real-time how reliable a firm’s AAA ratings are versus the average CRA.  With such a methodology we might be able to avoid both the government providing a misleading seal of approval as well as the damaging conflicts of interest between raters and issuers.  Issuers could still shop around, however if they choose to use a CRA that will give them a favorable rating but has a very low, public VORR they will likely suffer in the end as investors will be less likely to buy the security.  In fact, you would likely see the buy-side institute rules that incorporate VORR ratings into their buy decisions (i.e. no security may be purchased if rated by a CRA whose VORR is less than X).

Like I said, this is not an area I know particulary well.  Any readers at there more steeped in the CRA area that can offer some perspective.

I’ll assume this is my official 15 minutes (well, really 15 seconds) of fame.  Hopefully, the next time I show up I’ll either be the subject of a column or penning one.  Long way to go before that happens.  Back to work…