Statistics South Africa's Quarterly Labour Force Survey for the second quarter of 2026, embargoed until 11 August, put the number of unemployed people at 8,481,000 and the official unemployment rate at 33,6%, up from 32,7% in the first quarter. That quarterly move is the one the agency led on and the one every outlet printed: 345,000 more unemployed people in three months. It is also, on Stats SA's own arithmetic, real. Appendix 2 Table 2A of the same release gives the quarter-on-quarter rise a 95% confidence interval of 82,000 to 607,000 and a p-value of 0,01, and the 0,9 percentage-point move in the rate a p-value of 0,03. Measured against the threshold the document itself prints in section 8.6(iii) — below 0,01 highly significant, 0,01 to 0,05 significant, above 0,05 not significant — the quarter passes comfortably. Stats SA writes its decimals with a comma, and its figures are quoted here as printed.
The same PDF contains a second account of the same labour market, and the two do not agree. Table A on page 1 sets April to June 2026 against April to June 2025: 8,367,000 unemployed then, 8,481,000 now, a year-on-year rise of 114,000. Appendix 2 Table 2B prices that rise at a confidence interval running from minus 216,000 to plus 444,000, with a p-value of 0,50 — a coin flip. The year-on-year move in the rate, 0,4 percentage points, comes in at p 0,48. The number, the test that fails it and the threshold that defines the failure all sit in one 141-page document, on pages 1, 113 and 15 respectively.
Where the year's 114,000 came from is not narrated anywhere in the release, but it can be recovered by adding nine rows. Appendix 1 Table 2.3, on document pages 29 to 40, carries the unemployed count for each province across five quarters. Summed, the nine year-on-year changes come to plus 114,000 — the national figure, to the unit. One province supplies 263,000 of it: the Eastern Cape went from 956,000 unemployed in April to June 2025 to 1,219,000 a year later, a rise 2,3 times the size of the national total. The remaining eight net to minus 149,000, though that net conceals two rises. Gauteng added 52,000 unemployed and Mpumalanga 103,000, against six falls totalling 304,000, of which Limpopo's 101,000 is the largest.
Table 2.3B applies the agency's tests province by province, and three of the nine clear its own bar. The Eastern Cape's 263,000 comes with an interval of 135,000 to 390,000 and a p-value of 0,00; Mpumalanga's 103,000 scores 0,02; Limpopo's fall of 101,000 scores 0,03. The other six do not come close: Northern Cape 0,13, Western Cape 0,15, Free State 0,36, North West 0,44, Gauteng 0,52 and KwaZulu-Natal 0,58. The result is a comparison the release never makes about itself — a national year-on-year figure the agency's table cannot separate from zero, assembled out of provincial components three of which it can.
Stats SA published a second document on release day, and it settles which account the agency intends readers to take. The Statistician-General's 44-slide presentation opens on the quarter: “There were 345 thousand more people unemployed in Q2:2026 than in Q1:2026.” It carries no year-on-year change in unemployment anywhere in its 44 slides, and the figure 114 000 does not appear in it at all. Nothing is concealed — the annual comparison is a column in Table A on page 1, in front of everyone — but the agency's own narrative of its own survey runs on the quarter, and the year's number is left to the appendices and to whoever adds the columns.
The Eastern Cape figures themselves are not new, and the local press had them first. The Herald reported on 11 August that the province had added 13,000 jobs in the quarter, that its labour force had grown 160,000 to 2,565 million, and that it had shed 116,000 jobs over the past year. The Daily Dispatch followed on 12 August with the 263,000 year-on-year rise and the non-metro count of 918,000 unemployed at a 56,5% rate. What travelled alongside those numbers was an explanation, and the explanation used the quarter. Algoa FM's headline put it flatly: “Eastern Cape unemployment hits 47.5% despite 13 000 jobs added.” The DA's provincial statement the same day ran under “Eastern Cape unemployment surges as economy fails to absorb jobseekers.” The mechanism on offer was labour supply — discouraged people resuming the search and being counted as unemployed for the first time in a while — rather than work disappearing.
On the quarter, that reading holds up. Appendix Table 2.3A puts the Eastern Cape's fall in the “outside the labour force” category at 148,000, interval minus 254,000 to minus 42,000, p 0,01 — 71% of the entire national quarterly fall of 208,000 — while the province's employment gain of 13,000 carries a p-value of 0,72, indistinguishable from nothing. But 263,000 is a year-long number, and over the year Table 2.3B reverses the two rows exactly. Eastern Cape employment fell 116,000 across the twelve months, interval minus 211,000 to minus 21,000, p 0,02: the largest provincial employment loss in the country, and significant. The exit from outside the labour force over the same year, 99,000, scores p 0,17 and is not. The rise decomposes cleanly — 116,000 jobs lost, 99,000 people off the sidelines, 48,000 of working-age population growth, summing to 263,000 — and the province's own unemployment rate went from 39,5% to 47,5%, up 8,0 percentage points at p 0,00. The reassuring half of the story belongs to three months. The year's number is mostly jobs.
This is also the second consecutive print in which one province has moved the national figure, and the previous one ran the other way. The Q1:2026 QLFS, embargoed 12 May, showed national unemployment 91,000 lower year-on-year — p 0,59, also not significant — while the Eastern Cape was 183,000 higher at p 0,00 and the other eight provinces were 274,000 lower between them. Same province, same p-value, three months earlier, against a national fall rather than a national rise. Set the two releases side by side and they describe a national aggregate that has not moved measurably in a year and a single province that has moved sharply in both.
What changed between the two documents is what the Eastern Cape's rise is made of. In the Q1 release the province's year-on-year mechanism was the sidelines emptying: “outside the labour force” down 95,000, interval minus 182,000 to minus 8,000, p 0,03, with no statistically significant change in employment at p 0,43. By the Q2 release the two rows have swapped roles — employment loss significant at 0,02, the sidelines no longer significant at 0,17. In the space of one quarter, on the agency's own tests, the Eastern Cape stopped being a labour-supply story and became a job-destruction one.
Three checks argue against the provincial result being an artefact of the survey rather than the economy. The Eastern Cape's response rate, in Table G on document page 13, was 95,0% in both quarters, second only to Limpopo, so this is not thin sampling. The Eastern Cape back-series printed in the Q1 report — 956,000, 985,000, 1,016,000, 1,072,000 — is identical to the digit in the Q2 report, so it is not a revision working its way through. The caveat worth naming is structural rather than statistical: the QLFS master sample is still drawn on Census 2011 enumeration areas. The next print is not due until November, and the Reserve Bank's Monetary Policy Committee announces on 23 September, which makes this the standing labour-market reading into that meeting — a quarterly rise the agency's own table calls significant, wrapped around a year-on-year change it does not.