The record clay-plant hour that bought half-day Fridays
Showing the misleading chart
A trade newsletter we drew celebrates BLS’s output per hour for U.S. clay product plants (bricks, tile, pottery, sanitaryware, refractories): 113.6 in 2021 (2017 = 100), up a third in a year and the highest since the series began in 1987. Its committee works out that a 2017-sized year now needs 12% fewer hours and votes in half-day Fridays. But output in 2021 only climbed back to its 2019 level, while hours worked fell 17% below 2019, to 78.1 million. The record came mostly from the denominator. In 2022 the hours came back to 90.2 million and the ratio fell to 95.8, below 2017.
01The claim
The Bureau of Labor Statistics measures how much America’s clay product plants (brickworks, tile plants, potteries, sanitaryware and refractory makers) turn out per hour worked. In 2021 the index hit 113.6 (2017 = 100): up 33% on 2020’s 85.3, 13.6% above 2017, and the highest reading since the series began in 1987. If every hour now makes 13.6% more, a year as busy as 2017 needs only 78.8 million hours of work instead of 89.5 million, 12% fewer. A newsletter committee takes that at face value and votes in half-day Fridays.
02The trick
Every figure is BLS’s, the axis starts at zero, and the arithmetic is right. What goes wrong is reading a fraction as a property of each hour. Output per hour is real output divided by hours worked, and in 2021 the bottom of that fraction moved far more than the top. Output recovered from the 2020 slump, but only to 99.1, barely above 2019’s 98.2. Hours worked fell to 78.1 million, 17% below 2019’s 93.7 million and the fewest in the series up to then; BLS’s job count for the industry dropped to 36,500, from 42,100 in 2020 and 44,300 in 2019. A flat top over a shrinking bottom makes a record without anyone working any faster. The figures can’t say why the hours fell, whether plants were short of staff, ran lean, or the survey counts wobbled, and the ratio inherits the dip whichever it was. Neither of the usual culling stories, the weakest plants closing or old output still arriving, can be confirmed here; the lesson is the arithmetic of the fraction. None of that carries over once the hours come back, and in 2022 they did: 90.2 million, more than 2017’s 89.5, for 3.5% less output than 2017, and output per hour fell to 95.8. The same pattern runs through the whole recent series. The three highest readings, 107.2 in 2018, 113.6 in 2021 and 108.4 in 2023, each came in a year when hours sank to a new low, while output in those years (101.8, 99.1, 94.4) was no better than in 2013–14, and never above 2014’s 102.9. (Both drawings are ours; the newsletter, its committee and the motion are invented. Every figure is from BLS’s 2026 release, so it may differ from what a reader would have seen in 2022–23.)
03The fix
Draw the fraction apart. The redraw stacks three aligned panels on the same 1987–2025 time axis, all zero-based: real output, hours worked, and their ratio, with 2021 marked. Output runs close to 100 from 2010 to 2022, apart from the 2020 slump to 84. Since 2010 hours have wobbled between about 78 and 104 million. The ratio’s three highest points all land on years when the hours hit a new low, and 2021’s record is gone a year later, when the hours return. The general rule for any per-unit figure, whether per hour, per worker, per store or per rig: plot the numerator and denominator next to the ratio, and treat a record set in the year the denominator fell as news about the denominator first. Before planning on a per-unit gain, check whether it survives the count recovering.