I tend to disagree: unemployment can change just by the government giving out invalidity pension to previously "unemployed" people (famously in the UK) so this kind of long run government process completely messes the value of the unemployment time serie for a given country without any "structural" change in the economy.

So unemployment has no value for:

• comparing countries
  
• analysing time variation in one country
  
• describing the state of the job market in one country

Not far from useless :).

Let me quote Dean Baker (it's on my blog with links):

    Laurent,

    you are absolutely on the mark. In fact, many labor economists and increasingly macroeconomists are now arguing that the employment rate is the better measure since "unemployment," the state of not working, but actively looking for work, is ill-defined.

    As I noted before, using employment to population rations (EPOPS), the U.S. does not look better than old Europe, at least for prime age workers (source CEPR "Old Europe Goes to Work: Rising Employment Rates in the European Union").

    I should also point out that the U.S. data likely overstates EPOPs by 1.0-2.0 percentage points because people who are not employed are less likely to answer the survey (source CEPR "The Impact of Undercounting in the Current Population Survey").

    I give the media a bit more slack on using EPOPs as opposed to unemployment rates, because the switch in emphasis among economists is relatively new. On the other hand, there is no excuse for misrepresented a statistic, as the media does when it fails to note that the German unemployment rate is calculated with a different methodology.