Korea's construction deaths fell 19.3% on a workforce that shrank 3%

Employment on Korean construction sites in July was 1.865 million, 3.0% lower than a year earlier ^[6]. Run that against the 19.3% fall in first-half deaths ^[1] and the fatality rate comes out near 13.0 per 100,000 construction workers this year against 15.7 last year, a fall of about 17% ^[3]. The smaller payroll accounts for roughly two and a half points of the nineteen ^[10].

The injury series went the other way. Last year's first-half figure was near 16,417, so about 2,659 more workers were hurt this year ^[1]. Per 100,000 workers, injuries rose about 20% ^[4].

Fewer sites with fewer people should pull both series down. Deaths per recorded injury instead fell from one in 54.5 to one in 78.5, about 30% ^[2]. Either severity genuinely improved, or minor injuries that went unlogged last year are being reported now under inspection pressure, which shows up in paperwork before it shows up in risk. The figures given to Rep. Kim Wi-sang's office break out illness and accident deaths only for this year's first half ^[8][13].

Where the slump is unambiguous is the order book. Construction orders fell 30.6% from a year earlier to 15.8 trillion won in July, according to the Construction and Economy Research Institute of Korea's monthly market report ^[5]. That is one month's intake. Orders precede site work by months, and most of the labour contraction they imply has not reached the death count.

Construction was 23.2% of all first-half workplace deaths, the highest of any industry ^[4], and 58 of the 71 fewer deaths recorded across all industries came out of construction, 82% of the national decline ^[5]. The second half of last year carried 268 of the annual total, and this year's second half can reach 325 and still come in lower, about 21% above ^[6].

Analysts quoted by Seoul Economic Daily credit both the government's push to cut industrial accidents and a slump that reduced sites and workers ^[9]. In my view the first-half fall is carried more by enforcement and by reporting behaviour than by the missing workers, and the exposure decline the order book implies will show up in next year's figures. Employment is a headcount: an hours-worked series showing site exposure down 15% while payrolls fell 3% would break that reading.