For cartograms, certain attribute values are mapped to the size of regions, distorting the map.
Another key aspect is that the viewer of cartograms has to relate the distorted areas of the map to the original map locations in order to fully comprehend the effect.
This throws up the question what one should assume as the basic map, as we have different ways of projecting the globe onto a map. So there would be quite a big difference if we base the cartogram off of an equirectangular projection or an albers equal-area conic projection.
Does this mean one should state which basic map projection is assumed? Or is there a specific map type that is usually used as a basic or the differences are so small that it would not matter either way?