Lie Detection: How Can Financial Analysts Improve Their Ability to Discern the Truth?
In the finance profession the ability to suss out the truth is of critical importance. From analysts needing to assess the veracity of statements by corporate management to private wealth managers needing to understand their clients’ actual financial motives, the ability to tell fact from fiction is highly prized.
In “Why Do Lie-Catchers Fail: A Lens Model Meta-Analysis of Human Lie Judgments,” a research paper published in the American Psychological Association’s Psychological Bulletin, Maria Hartwig and Charles F. Bond, Jr., provide valuable insight into the behavior of lying. “Meta-analysis” is the compilation of many studies into one overarching data set from which more sweeping conclusions can be reached. What follows is a compilation of more than one hundred studies conducted over 50 years about lying and lie detection.
The corpus of research on lying has demonstrated that people are generally very bad at detecting lies. In fact, Hartwig and Bond’s meta-analysis shows that generally people are able to detect lies about 54% of the time. This is only slightly better than chance. Not only amateurs have trouble sussing out a lie; the authors note that “contrary to common expectations, presumed lie experts who routinely assess credibility in their professional life do not perform better than lay judges do.” Why is this?
In the psychological community, two main theories have been put forth to help explain the poor ability of people to detect lies:
- People have a false stereotype about what constitutes lying behavior. That is, they are using the wrong cues to detect lies.
- There is only a minute behavioral difference between truth-tellers and liars.
Hartwig and Bond tested these hypotheses using meta-analysis, and found that the first hypothesis did not hold. In fact, the researchers found that generally people are using behavioral cues to identify lying. However, the researchers also found that the behavioral cues that subjects reported using are not the cues they actually utilized to detect lies. In other words, the subjects were, able to detect lies at an intuitive level, but they didn’t consciously know how they were doing it.
As for the second hypothesis, the researchers did find strong evidence that there are only small behavioral differences between truth-tellers and liars.
What about the overall accuracy of truth detectors? Hartwig and Bond measured the ability of truth detectors as a Pearson product-moment correlation coefficient — that is, the correlation between actual lying (RDec), perceptions of lying (RPer), and the accuracy of the detection methods (G). Put mathematically, the accuracy of truth detection is calculated as follows:
racc = RDec × RPer × G
0.21 = 0.36 × 0.63 × 0.93
The accuracy of lie detection is hurt most by the lack of valid behavioral cues (r = 0.36). That’s another way of saying that the behavioral difference between truth-tellers and liars is small. In general, lie detectors’ perceptions of lying behavior are strong (r = 0.63). Finally, the lie detectors appear to be using the correct cues to detect lying (r = 0.93).
In different assessments of human judgment of behaviors other than lying, the average accuracy coefficient is much higher. One study of peoples’ ability to perceive other qualities in human behavior had an accuracy coefficient of 0.56. This compares to the accuracy coefficient of 0.21 for lie detectors .
What Cues Can I Rely On?
Most importantly, the best indications of a lie are not single behaviors from the liar but the overall impression the liar makes on the truth detector. In fact, research has continually demonstrated that overall impressions of lying trump individual cues.
Below are the top five behavioral cues to deception, each of which has statistical significance of at least 95%. Though each of the following behaviors is positively correlated with lying, note the low correlations to lying. In other words, just because you witness the following behaviors does not necessarily mean that the speaker is lying.
Cue to Deception | Actual Correlation to Lying |
---|---|
Indifferent (speaker seems unconcerned) |
0.45 |
Thinking hard |
0.29 |
Ambivalent (communication seems internally inconsistent or discrepant) |
0.19 |
Not spontaneous (statement seems planned or rehearsed) |
0.19 |
Not fluent (miscellaneous speech disturbances) |
0.19 |
By far, the strongest indication that someone is lying is indifference. Yet indifference is a sign of lying less than half of the time (r = 0.45). The impression that a speaker is thinking hard is also a relatively strong indication that the speaker is lying (r = 0.29). After these first two behaviors, the correlation coefficients all drop in value. So ambivalence and lack of spontaneity and fluency are signs of lying, but weak signs.
What Cues Are Less Reliable?
Here are the behavioral cues that people reportedly use for lie detection but that have proved unreliable, all with a significance of 95% or higher.
(a) Cue to Deception | (b) Actual Correlation to Lying | (c) Perceived Correlation to Lying | (d = b – c) Absolute Error |
---|---|---|---|
Competency |
-0.02 |
-0.59 |
0.57 |
Arm movements |
-0.19 |
0.37 |
0.56 |
Level of involvement |
0.05 |
-0.42 |
0.47 |
Pleasant face |
-0.05 |
-0.44 |
0.39 |
Plausibility |
-0.11 |
-0.47 |
0.36 |
As the above list shows, people tend to think that someone they perceive as incompetent (negative correlation to competence) is lying; yet the actual correlation between perceived incompetence and lying is almost zero. Additionally, if communicators use lots of arm movements, people tend to think it is a sign of lying (positive correlation of 0.37). But, in actuality, a person who moves his/her arms slightly less (negative correlation of -0.19) is more likely to be lying. Lie detectors tend to think that a lack of involvement is a sign of lying, but in fact, to a slight degree (0.05) the more involvement, the greater the chance that a lie has been told. People also strongly feel that an unpleasant face (r = -0.44) is a sign of lying, when in fact, the chance of this is only slightly higher than zero. Perhaps most surprisingly, implausibility is actually not a strong indication that a lie is being perpetrated.
Sometimes truth detectors err by using the proper criteria to detect a lie but placing too much significance on a particular cue. Here then are the top five examples of that error, all with significance of 95% or higher:
(a) Cue to Deception | (b) Actual Correlation to Lying | (c) Perceived Correlation to Lying | (d = b – c) Error |
---|---|---|---|
Ambivalent (communication seems internally inconsistent or discrepant) |
0.19 |
0.49 |
0.30 |
Not spontaneous (statement seems planned or rehearsed) |
0.19 |
0.48 |
0.29 |
Vocal uncertainty (impressions of uncertainty and insecurity, lack of assertiveness) |
0.14 |
0.43 |
0.29 |
Unfilled pauses (periods of silence) |
0.01 |
0.27 |
0.26 |
Gaze aversion |
0.05 |
0.28 |
0.23 |
How Can I Improve My Ability to Detect a Lie?
Most importantly, the authors suggest a holistic approach to lie detection. That is, do not rely upon individual behavioral cues, as a preponderance of lying behaviors is more indicative than any single cue.
Another method for improving your ability to detect a lie is to trust your intuition rather than what you perceive are good behavioral cues. Studies continually demonstrate that when lies are successfully detected the methods of detection ascribed by detectors are not the ones they actually use. This suggests a disconnect between the potency of unconscious detection and impotency of conscious method.
The typical prescription of truth-detection trainers is to give prospective truth detectors a list of behavioral cues to look for and then give feedback on performance to improve the results. But these methods have statistically been demonstrated to be ineffective, or only marginally effective.
The authors of the “Why Do Lie-Catchers Fail?” feel that the best way to improve one’s ability to assess a lie is to increase the difference in the behaviors of liars and truth-tellers. One way to do this is to engage in interactional interviews. Because lying requires greater cognitive energy than telling the truth, you can increase the cognitive demand of your questions. For example, ask someone a question that challenges them to place a detail of their complex story back in its proper chronological context to see if they can remember where the detail fits in the timeline.
In conclusion, the business of investing requires an ability to discern the truth and veracity of the information you are using in your analytical process. Yet, most of us are in fact very poor at catching lies when they are told. Statistically, more than 50 years of research has shown this is because there is not that much difference between liars and truth tellers in how they communicate, and because, in all likelihood, you are ignoring your intuitive faculties.
To catch a liar (lawbreaker, cheater, villain…etc), you need to think like a crook. Does this mean the individual good at finding liars is just as bad as a crook? Also, why would anyone think about hiring an individual supposing good at identifying liars when there is a chance the same individual could cheat the employer that employees that individual. In my opinion, the answer rest on the person making the decision for the department / firm.
Hello Tyc,
If I may borrow from a great tradition of thinkers, all willful events have elements of thought, word and deed. Thought in the example of lying would be thinking like a criminal. Word would be expressing a commitment to do the criminal act. Deed, of course, is actually doing the criminal act. The theoretical person who is good at lie detection does not engage in the final two stages of the willful criminal act, just the first part, the thought part of the thought-word-deed chain.
With smiles,
Jason
Jason,
Another excellent article, and very easy to scan (thanks to your tables, subheads, and bullet points).
The two key takeaways I got were:
1) Holistic approach is best–trust your instincts
2) Beware overconfidence! As you noted, even reliable behavioral cues were predictive less than 50% of the time.
Thanks for a great article on a difficult subject.
Rob
Hi Robert,
Thank you for your kind words. I am committed to providing more of this type of content for our audience. This is because, as an ex-practitioner, I know that evaluating the veracity of the truthfulness of a person’s statements is simutaneously important and difficult. You summarized the take-aways very well.
With smiles!
Jason