The resignation wave that was measured from the bottom of the hole

The rebased index The cherry-picked window

Dark navy slide, masthead “People analytics · Quarterly retention briefing · Q3 2026 · US labour market”, headlined “US quits are still 54% above base. The resignation wave hasn’t broken.” A standfirst says the line is total nonfarm quits from the BLS JOLTS survey, seasonally adjusted, every month from April 2020 to the latest release, nothing smoothed, indexed to April 2020 = 100; quits more than doubled to a peak of 226 in April 2022 and still stand at 154. The chart, labelled “US quits · monthly · index, Apr 2020 = 100”, runs from April 2020 to August 2026, with ticks at 2021–2026, on an axis from 80 to 240, with a dashed amber line at 100 marked “Base: April 2020 = 100”. A teal line starts on the dashed line at the left edge, climbs steeply through 2020 and 2021 to a dot labelled “Peak 226 · Apr 2022”, then eases down through 2023 and 2024 and wanders between about 150 and 175 to a final dot labelled “154 · Aug 2026”. Cards on the right read “+54%, US quits versus base · August 2026, peak +126% (April 2022)”; a read-out saying the wave has crested but not broken, with quits at 2.26 times base at the peak and 1.54 times base today; a recommendation to hold the FY22 retention and counter-offer budget; “Why index”, saying raw headcounts run into the millions and hide the trend, and that because the base is the start of the window “above” always means “up since then”; and “What we did not do”: no smoothing, no months dropped, no second axis. A footnote credits BLS JOLTS via FRED series JTSQUL, retrieved 16 October 2026, notes August 2026 is preliminary, and says the slide is a demonstration drawn by Misleading Charts in the manner of an HR briefing, with the people team, read-out and recommendation invented and every figure real. — Showing the misleading chart

Cream page, kicker “The same series, not indexed · US quits, December 2000 – August 2026 · quits per month”, headlined “Measured from the trough, quits are up 54%. Measured from the 2019 average, they are down 13%.” A standfirst says the briefing’s window opens on April 2020, the lowest month for quits in seven and a half years, so its first point and therefore its 100 is the bottom of the pandemic collapse, and that drawn as actual quits over the whole published series the line has been below its 2019 average every month since December 2023. The main chart is a thin black line of monthly quits on an axis from 0 to 5 million, December 2000 to August 2026. It dips from about 3 million in 2001 to about 2.2 million in 2003, recovers to about 3 million by 2006–07, then falls to a dot labelled “1.56m · Aug 2009, series low”, climbs steadily through the 2010s to about a dashed reference line marked “2019 average · 3.51m”, drops vertically to a red dot labelled “1.99m · Apr 2020, the briefing’s base”, shoots up to a dot labelled “4.50m · Apr 2022, series high”, and slides back to end below the 2019 line at a dot labelled “3.07m, Aug 2026”. The stretch from April 2020 onward is shaded and labelled “the briefing’s window”. To the right, a dot plot headed “One reading, four bases: August 2026 quits as an index, base = 100” shows four bars against a dashed line at 100: Apr 2020, the trough, 154, in red; Dec 2000, series start, 106; 2019 average, 87; Apr 2022, the peak, 68. A note headed “What rebasing can’t move” gives August 2026 as 3,066,000 quits, 1.9% of employment, the 2019 average as 3,513,000 and 2.3%, the April 2022 peak as 4,499,000 and 3.0%, and says the rate has sat at 1.9–2.0% for the past twelve months. A note headed “Quits are back below 2019” says thirty-three months in a row under the 2019 average, and that the pressure the briefing reads as “still elevated” is mostly the depth of the hole it measured from. A footnote gives the source, FRED series JTSQUL and JTSQUR retrieved 16 October 2026, notes that August 2026 is preliminary and recent months are subject to revision, lists the four base values used, and says both drawings are ours, the briefing invented and every figure real. — Showing the misleading chart

Be the first to star this exhibit

A retention briefing we drew from BLS data indexes US monthly quits to April 2020 = 100 and finds them still 54% above base six years on. April 2020 is the first month on the chart, which is where the textbook says to put the base, and it is the lowest month for quits in seven and a half years. Against the 2019 average, the same August 2026 reading is 13% down, and quits have been below that average for 33 months running.

US Bureau of Labor Statistics, Job Openings and Labor Turnover Survey (JOLTS): total nonfarm quits, thousands, seasonally adjusted, and quits rate, via FRED series JTSQUL and JTSQUR, monthly, December 2000 – August 2026 (August 2026 preliminary), retrieved 16 October 2026. The briefing, its read-out and both drawings are ours., 2026Added October 16, 2026

01The claim

Six years after the pandemic shutdowns, Americans are still quitting their jobs at a pace well above normal, so retention budgets should stay where they were at the height of the Great Resignation.

02The trick

Nothing on the slide is bent. Every month is plotted, nothing is smoothed, and the index is calculated correctly. The base even follows the textbook advice, sitting at the start of the window so that “above” always means “grew since then”. The catch is which month the window starts on. Rebasing divides the whole line by its value in one month, and the usual habit of basing at the first month plotted ties the two choices together: choosing where the chart starts also chooses the 100. This chart starts in April 2020, the month the pandemic shutdowns cut quits by a third in four weeks. That month 1.99 million quits were recorded, the fewest since September 2012 and 43% below 2019’s average of 3.51 million a month. Dividing by the bottom of a collapse makes everything after it look high. August 2026’s 3.07 million quits come out at 154, and the slide reads that as “still 54% above base”, which is arithmetically true and says nothing about whether quitting is unusually common. The index has no level on it, so the one comparison that would answer the question, now against before, never appears. The read-out does the rest: “the wave has crested but not broken” is a sentence about the divisor. (The briefing, the people team and the recommendation are our invention; every number is the Bureau of Labor Statistics’.)

03The fix

Draw it as actual quits, over the whole published series, and the briefing’s window turns out to be a short stretch that starts at the bottom of the plunge. The only months since 2000 with fewer quits fall in the long slump after the 2008 crash, which bottomed at 1.56 million in August 2009. The 2021–22 surge is real: quits reached 4.50 million in April 2022, the series high, with a quits rate of 3.0% of employment. But the line had dropped below its 2019 average by December 2023 and has stayed there for 33 straight months. August 2026 is 3.07 million, a rate of 1.9%, against 3.51 million and 2.3% in 2019. As a rate, which allows for a bigger workforce, 1.9% is below even December 2000’s 2.2%. As an index, that one August reading is 154 based on April 2020, 106 based on December 2000, 87 based on the 2019 average and 68 based on the April 2022 peak. The month in the divisor decides whether the same number reads as a surge or a slump. When a chart is indexed, put the base in the caption and choose it for a reason you could give before seeing the line. “The last normal year” is a reason; “the first month in the export” is not. Then keep one fact in words that no base can move: “3.07 million quits in August 2026, against 3.51 million a month in 2019”.