Left skewed data with a long tail.  Maxwell-Boltzmann or Landau distributions might model this well. 

I suspect 20s and 30s are multi model and smoothed out later.

Would be cool to see the full data set