Relearning Statistics Can Reduce Your Risk of Manipulation

When we read that 50% of people believe this or 75% of people are worried about that, it provides a level of reassurance as to the validity of what we are reading. Big percentages make us feel psychologically safe, and our brains instinctively feel secure in the knowledge that we won’t be alone in our beliefs. The problem is that while a statistic may be true, the very point that they represent a true-false relationship makes them inherently flawed and leaves us susceptible to both unsound logic and frequent manipulation. Statistics are one of the reasons that propaganda is so effective, and with the rise in misinformation across social media platforms, they leave us vulnerable to fake news and hoaxes too.

Statistics are often viewed with reverence, seen as an irrefutable ‘fact’ that is numerically verified and isn’t based on opinion, thereby validating the point that’s being made. The media love them for straightforwardly illustrating a key point, but so do people working in advertising and sales, and those who seek to manipulate you. It’s why understanding statistics is an essential tool for inoculating your thinking against potential bad actors, and assuring your continued ability to think critically and independently.

By definition, statistics are listed as “a fact or piece of data obtained from a study of a large quantity of numerical data”; a fact is in turn defined as “a thing that is known or proven to be true”. It is this that gives them the perception of irrefutable truth. Statistics offer an inherently true-false or yes-no position on a subject. The hitch is that the majority of the populace doesn’t fully understand statistics – how they are calculated, what they represent, and what their flaws are – leaving them with a yes-no bias when reading what they portray. The big problem is that statistics, in true terms, are representative only within their context, and without context are typically rendered entirely useless.

Obtaining context

There are two key problems with how we interpret statistics. The first is that they depend on the context in which they are collected, and the second is that they are open to multiple interpretations of the same results. One set of numerical results can produce multiple “correct” statements, some of which will conflict with each other. If we take a random sample from a bag of sweets for example, and the sample contains four green sweets, two red sweets, and four half-red, half-green sweets, then the following statements are all technically correct:

40% of these sweets are green
20% of these sweets are red
60% of these sweets are not entirely green
60% of these sweets are not entirely red
60% of these sweets contain the colour red
80% of these sweets contain the colour green
None (0%) of these sweets are yellow.

Every one of these seven statements is ‘true’ offering a ‘fact’ about the sample from the bag of sweets, but each one communicates something completely different and unique to the interpretation. What’s more, the lack of context provides a lack of supporting information, such as how many sweets were originally in the bag; whether any sweets were eaten prior to the experiment; and whether there are any other coloured sweets in the bag – any of which could change the validity of the statements when viewed through the lens of context.

This is a very simplistic example, based entirely on numeric values – items that are easily counted. For the statistics that we read in the media, both the ‘fact’ and the context are inherently more complicated, presenting correlations between two factors, increases and decreases over time, or information that is factually true in one context, but does not account for information outside of this context.

An example is the studies that perpetuate the mantra that “breast is best” when it comes to feeding a baby, directly correlating a causal link between breastfeeding and overall intelligence and performance at school. Persistent historic studies identified a positive correlation between the two factors, implying a direct causal link, but subsequent analysis has highlighted flaws in these early studies. Scientists had originally assumed that because one increased as the other increased, there must be a direct causal link, but what these studies fail to account for is the relationship between the type of household and the likelihood to breastfeed. In fact, later studies have discounted the link between breastfeeding and intelligence/academic performance, in favour of the relationship between socio-economic background and academic performance. These later studies instead found that on average, children with better performance at school, typically have parents at home who have achieved higher academic standards themselves, or have secured an overall higher household income, and therefore have the luxury of more time and skills to invest in their children and more money with which to pursue supportive activities. They also found a positive correlation between higher socio-economic status and incidents of breastfeeding, because those households with higher income could afford to take longer maternity leave, and could afford to breastfeed their children for longer. A correlation, not a causation. Indeed, many studies have since found no difference in the performance of children from the same socio-economic background, regardless of whether they were fed breastmilk or formula. Yet still the myth perpetuates.

During the Covid-19 pandemic, the vaccination debate provided further evidence of the dangers of statistics, and the generally poor standard of interpretation we have across the populace. Months after vaccinations first started, evidence emerged of a potential causal link between the Covid-19 vaccination and an increased risk of thrombocytopenia (blood clots) in the weeks following the vaccination. The media began reporting, and initial studies emerged evaluating the risk to the general populace. The challenge is that the studies found a 30% increased risk of blood clots after a first dose of Oxford-AstraZeneca compared with Pfizer-BioNTech. This was interpreted by many as a 30% risk of a blood clot, or almost one in every three people. This is not however what the statement means. First, it is a 30% increase when comparing one manufacturer with another, and second, 30% does not represent the absolute risk of developing a blood clot. A large percentage of a very small number is still a small number; in this case, the absolute risk was <0.001% or less than 1 in 1,000 which is significantly lower than the perceived one in three risk. By January 2022, over 10 billion COVID-19 vaccinations had been administered, with very few overall incidents of blood clots. The damage was done however; a poll from the U.S. in April 2021 revealed that 76% were “very or somewhat concerned” about serious side effects from vaccination and 70% felt that “the COVID-19 vaccines are not as safe as they are said to be.”

Modern media, particularly social media platforms, have become a key source of information and news for much of the population. But these same platforms have also supported a maelstrom of fake news, misinformation and even deliberate disinformation campaigns. Many of these “news” stories rely on statistics, and big, scary sounding numbers to create widespread panic and manipulation of the general populace. A study by Oxford University found that social media manipulation campaigns are widespread across all countries, and that a key proponent in the spread of misinformation is the use of “citizen influencers” when well-meaning individuals spread unverified research and communications.

Reducing your susceptibility

The key to reducing your susceptibility is to re-educate yourself on statistics including what they say, but also what they don’t say. For starters, always go back to the original source being quoted by the numbers, and analyse the quality of the study and the accuracy of the reporting. Next determine whether the original source can be trusted, and determine whether there can be any other explanation for the results from the study. Finally, seek alternative sources; similar academic studies and conflicting ones, to help determine whether there is merit in what’s being represented.