Rounding and Appropriate Precision in Reported Data

A news report saying the average family has 2.3 children isn't claiming any family has a fractional kid. It's reporting an average, a sum divided by a count, and averages don't have to land on whole numbers even when every individual case does. The .3 is honest information about the whole group, not a measurement of one family.

Separately, when a single number is rounded, that rounding sets a boundary on what the true value could have been. If a population is reported as 4.2 million rounded to the nearest hundred thousand, the true value could be anything from 4.15 million up to (but not including) 4.25 million. Writing 4.200000 million would look more precise without being any more true. Precision is about how fine the rounding unit was, not how many digits you bother to write down.

This connects to two-way tables and percentage breakdowns elsewhere in data description. When several related values, like the rows of a percentage table, are each rounded independently before being reported, their individual rounding errors don't cancel out. Each row can be off by up to half a percentage point, and those small errors can stack so the column adds to 99 or 101 instead of 100. Recognizing that pattern means you don't panic over a table that's 'off' by a point or two, and you don't try to force every rounded value to be exact when the real quantity was fractional all along.