This is surprising given their laws making it difficult to fire existing workers. But there's another way to look at the data. Suppose a firm is pretty good at estimating worker productivity which occur at low (L), medium (M), and high (H). But the firm is not perfect and sometimes one level off: L can be mistaken for M, H can be mistaken for M, M can be mistaken for L or H, etc. If you know that getting an L will lock you into that person, and since the Ls can make the Ms and Hs worse, you refuse to hire anyone that's an L or an M just be sure. Therefore, French companies are chuck full of Ms and Hs while the unemployed are all Ls and Ms. There may not be two Americas, but there just might be two Frances.