Here is a number that should worry you, and then here is why it shouldn't.

An ONS labour market rates series, aggregated across economic activity, age group and seasonal adjustment, reads 35,920.3 in 2020. The following year it reads 11,933.5. That is a fall of 23,986.8, or roughly 67%, in twelve months (Source: ONS, Labour market overview). Two-thirds of the thing, gone.

If that were real, it would be the single largest labour market event in modern British history. It isn't real. It's the shape of a spreadsheet.

Look at what surrounds it. From 2012 to 2020, this series barely moves. The highest value is 36,218.8 in 2013, the lowest is 35,825.6 in 2019, and the entire nine-year spread is 393.2 points. That is about 1% of the total. Nine years covering austerity, Brexit, and a pandemic that shut the country down, and the line is flatter than a snooker table. The pandemic year of 2020 actually came in higher than 2019, by 94.7 points.

A series that ignores COVID cannot then be halved by 2021. Something else is going on, and the clue is at the top of the file: these values are summed across categories. When a category stops being collected, or an age band drops out, or one version of the seasonal adjustment goes missing, the total falls. Not because fewer people are working. Because fewer rows are being added up.

The other end of the file says the same thing. The first year, 2011, reads 15,115.7. The next year reads 36,217.6, a jump of 21,101.9. That is a 140% rise in a single year in a measure that then refuses to move for nearly a decade. Both 2011 and 2021 are partial years wearing the same clothes as the complete ones. 2011 carries about 42% of the 2012 total; 2021 carries about 33% of the 2020 total.

The honest alternative reading, and it matters: I cannot prove from this table alone that the 2021 figure is incomplete rather than genuinely different. A definition change, a discontinued sub-series, or a real methodological break would all produce the same footprint. What the table does prove is that the 2021 number is not comparable with the nine years above it, and that anyone quoting a 67% fall from this file is quoting an artefact of how the rows were combined.

My view, and it's only mine: partial years should never be published looking identical to whole ones. A single flag in the release, one column marking coverage, would stop this. Instead the burden lands on whoever downloads it, and most people downloading a government spreadsheet are not statisticians. They are a council analyst, a charity researcher, a journalist on deadline.

Note also the age of this. The most recent figure here is from 2021. Whatever the labour market has done since, this file cannot tell you.

The lesson is cheap and useful. When a number moves by two-thirds in a year and by 1% in the nine years before it, the story is almost never in the world. It's in the file.