Should We Have Equal or Unequal Sized Class Intervals?

Class intervals are usually kept equal, but unequal intervals are the better choice in two situations.

First — a very wide range. Variables like income can run from almost zero to hundreds of crores of rupees. Equal intervals then fail either way: keep them moderate and equal and you get an unmanageably large number of classes; make them large and you hide detail at the very low and very high ends of income.

Second — data heavily bunched in a small part of the range. If most observations crowd into a narrow band, equal-width classes there would lump many distinct values together and lose information.

In every other case, equal class intervals are used. (The mathematics-marks distribution is later re-done with unequal classes to show exactly this benefit.)