One reason why the graph has multiple trends is that its multiple time series dumped on top of each other (the article entirely glosses over it being data from different points in time which would potential exhibit independent trends in premature deaths).
The article does mention that "the outlying figures to the right are those for South Korea". If I'm interpreting that correctly, it's the neat little chain to the right hand side which shows South Korea's working hours rapidly declining and potential years of life lost also falling over time; but if working hours decline you'd expect a spurious correlation to appear even if there was no causal link since healthcare standards also steadily improve over time. To the extent there is a causal link between the two in South Korea - as opposed to coincidental secular trends over time - it's probably more down to economic development than stress.
I'd expect a publication called the Economist to know how to handle panel data...
The article does mention that "the outlying figures to the right are those for South Korea". If I'm interpreting that correctly, it's the neat little chain to the right hand side which shows South Korea's working hours rapidly declining and potential years of life lost also falling over time; but if working hours decline you'd expect a spurious correlation to appear even if there was no causal link since healthcare standards also steadily improve over time. To the extent there is a causal link between the two in South Korea - as opposed to coincidental secular trends over time - it's probably more down to economic development than stress.
I'd expect a publication called the Economist to know how to handle panel data...