Men lie. Numbers don't — if we let them speak

Every morning brings more information than any generation before us could have read in a lifetime. Alerts, feeds, clips, threads and charts arrive faster than we can weigh them. One might expect a society so well informed to see itself clearly. The opposite is happening: the more we are told, the harder it becomes to know what is actually going on.

The problem is not a shortage of information but a surplus of it, and the way that surplus is sorted before it reaches us. Social networks and recommendation engines do not select what is representative; they select what holds attention. What holds attention is the rare, the shocking, the extreme. The result is a magnifying glass permanently held over the edges of reality, while the middle, where most lives are actually lived, falls out of frame.

A magnifying glass on the extremes

Psychologists described the mechanism long before social media. In 1973, Amos Tversky and Daniel Kahneman showed that we judge how frequent an event is by how easily examples come to mind: the availability heuristic. A plane crash replayed a thousand times feels more common than the thousands of uneventful flights that left on time that same day. Feeds turn this quirk of the mind into a business model: the more vivid the example, the more it circulates, and the more it circulates, the more frequent it seems.

Falsehood has a further advantage: it travels faster than the truth. In the largest study of its kind, published in Science in 2018, Soroush Vosoughi, Deb Roy and Sinan Aral followed about 126,000 contested stories shared by some three million people on Twitter between 2006 and 2017. False stories were 70% more likely to be retweeted than true ones, and they spread farther, faster, deeper and more broadly. The authors were clear on the cause: it was people, not bots, who did most of the spreading. Novelty and emotion are rewarded; accuracy is not.

Generative AI now lowers the cost of producing such material to almost nothing. A convincing image, a plausible quote or a fabricated chart can be made in seconds and tailored to each audience. In its Global Risks Report, the World Economic Forum ranked misinformation and disinformation as the most severe short-term global risk in 2024 and again in 2025; in 2026 it still ranks second, behind only geoeconomic confrontation.

Why it matters: we decide together

This would be a private inconvenience if it stopped at our screens. It does not. Citizens who see the world through its extremes then vote, debate and make collective decisions. A society that overestimates rare dangers and overlooks common ones will spend its attention, its budgets and its political energy in the wrong places. A decision can be perfectly democratic and still rest on a distorted picture of the facts.

The question that matters for most public choices is not "what is the worst thing that happened?" but "what usually happens, to how many people, and is it getting better or worse?" Those questions have answers, and the answers are statistical.

The only antidote: method

There is no shortcut. The answer to too much information is not less information but better reading of it, and the best reading tool humanity has built is the scientific method and its instrument of measurement, statistics. Statistics replace the anecdote with the distribution, the striking case with the typical one, the impression with the order of magnitude.

Take a single example from the data on this site. Across the 193 countries assessed by the United Nations Development Programme, the Human Development Index ran in 2023 from 0.388 at the bottom to 0.972 at the top. Either extreme makes a headline. The median country stood at 0.762, and it is the median, not the extremes, that describes the condition of most of humanity. The same holds for life expectancy: in 2024 the median across 199 countries was 73.9 years, a figure that no single tragedy and no single record can tell you on its own.

Numbers can be made to lie too

It would be naive to stop there. The same tool that corrects bias can be used to manufacture it. Statistics do not lie by themselves, but they can be selected, cut, framed and timed until they say what someone wanted to hear.

The most common distortion is not the false number but the true number chosen with care: the one year that supports the argument, the one country that fits the story, the one indicator that flatters the thesis. This is confirmation bias dressed in data. The economist Ronald Coase warned against it decades ago, in a line later popularised as: "If you torture the data long enough, it will confess." The figure is real; the conclusion is not.

The traps we watch for

Mastering the tool means knowing where it breaks. These are the traps the Gazette checks for before any figure is published.

Noise. Every series moves a little from year to year for reasons that mean nothing: sampling, revisions, weather, a calendar effect. A rise of a few tenths is often not a trend. We look at several years and at the size of the usual fluctuation before calling anything a change.

The base effect. A percentage change depends on its starting point. After a deep fall, a modest recovery can look like a spectacular jump; after an exceptional year, a return to normal can look like a collapse. We always show the level as well as the rate.

The threshold effect. When many people sit just below or just above a line, a poverty line, an age limit, a tax bracket, a tiny move can produce a large change in the headline count while almost nothing has changed in reality. We say where the threshold is and how many people sit near it.

Averages that hide distributions. A mean can rise while most people are no better off, if the gains go to a few. We report medians, ranges and inequality measures alongside averages.

Simpson's paradox. A trend that holds in every group can reverse when the groups are pooled. In a well-known 1975 study of graduate admissions at the University of California, Berkeley, men appeared to be admitted at a higher rate overall, yet within most departments the gap disappeared or favoured women; the difference came from which departments each group applied to. Aggregation can mislead as surely as omission.

Survivorship and selection. During the Second World War, the statistician Abraham Wald pointed out that the bullet holes on returning aircraft showed where a plane could be hit and still fly home; the armour belonged where the surviving planes showed no damage. Data that only includes the survivors tells the wrong story.

Correlation and causation. Two series that move together may share a third cause, or none at all. We describe associations as associations.

Projections. A forecast is not an observation. On Statlas, projections are always labelled and drawn apart from measured data, and the Gazette follows the same rule.

Our line: neutral, sourced, non-partisan

The Gazette exists to apply this discipline to the questions our societies argue about. Our method is simple to state and demanding to follow. We start from the question, not from the conclusion. We use one identified source per indicator and link every figure back to it. We show the typical case before the exceptional one. We publish the base, the threshold and the uncertainty with the number. And we publish what contradicts our expectations as readily as what confirms them.

We take no side but the evidence. Our aim is not to win an argument but to give every side the same facts to argue from, so that collective decisions can be made rationally rather than by whoever shouts the most striking exception.

Men lie; numbers do not, provided we let them speak in full: with their source, their context and their limits. That is the promise of this Gazette.