Recently, the journal published by an American business school carried an article with the intriguing title, If You Say Something Is "Likely," How Likely Do People Think It Is? The subject of the article was the imprecision of the words that are used to convey the likelihood (or otherwise) of some event or outcome, and how that can be corrected. Since promises and expectations lie at the heart of savings and investment choices that we make, I found the study fascinating, but some of the suggestions a little suspect.
What makes the article fascinating is that the authors, one of whom is a data scientist at Twitter and the other a financial researcher/academic, conducted open research on exactly what people thought was being said when such words were used. Some examples of the words and phrases they tested were Always, Certainly, Almost Certainly, With High Probability, Usually, Often, Serious Possibility, With Low Probability, Rarely and Never. In the study, each subject is asked to give the exact percentage that they feel that each word or phrase corresponds to. So if you feel 'Certainly' means 90% then you fill that in and so on.
As one would expect, there are some words which mean almost the same to everyone while others have far more ambiguous meanings. Even so, the spread of the results is surprisingly wide. I would have thought 'Always' would mean 100% for everyone, and even though 75% of the respondents were at 98% or above, the 5th percentile of the response was as low as 90%. 'Certainly' was spread from 80% to 100%, and 'Almost certainly' was the same.
Unsurprisingly, the most ambiguous terms were in the middle of the range. However, the wide spread of probabilities that these terms mean to different people look quite problematic. One of the widest spreads is for 'Serious Possibility', which ranges from 30% to 90% at the 5th and 95th percentiles, with even the 25th and 75th percentiles at 45% and 80%. 'Real Possibility' is almost as bad, although a little lower all along. I was truly surprised at this, having thought both to be 60%.
What does this mean in practice? It means that we should realise that phrases such as 'Real Possibility' and 'Serious Possibility', are meaningless. Or actually, they are worse than meaningless because they hold wildly different meanings for different people. 'Likely', 'Frequently', 'Probably' and many more have quite ambiguous meanings too. If you read or listen to business or investing news, or investment advice, you will come across these words all the time. Analysts and researchers are always saying that some investment is very likely to do well or that there is a real possibility that interest rates will fall, or that there is a serious possibility that the markets will stay stagnant. Now you know that they could mean anything, and therefore actually mean nothing.
However, there's a further problem in what most people (including the authors) suggest as a solution, which is that everyone should give probabilities. Probabilities, as used by most talking heads, leave even bigger escape routes. I remember, a few months ago, just before an RBI interest rates announcement, a prominent economist saying on TV that he thought there was an 80% probability that rates would stay unchanged. When the central bank actually dropped rates, he promptly pointed out that according to his estimates, there was a 20% chance of interest rates changing. Basically, that means that a sufficiently smooth operator can never be wrong when quoting probability percentages! Heads I win, tails you lose.
In the original article, the cure for this is that forecasters' track records be tabulated and a proper success score be calculated using a widely-recognised method. However, this kind of a suggestion is not much use when you are faced with, say, a financial salesman who is promising that an investment will definitely get you high returns. The only thing to do is to pin people down on percentages and numeric promises. And be aware that the same words mean different things to different people. And do try the survey yourself at www.probabilitysurvey.com.