For some unexplained reason, Rosling sets GDP on a logarithmic scale. This basically normalizes the assumption that wealth is extremely unequal (i.e., follows some sort of exponential curve). Which is not to say that this isn't the natural state of things, perhaps it is, but it's misleading to adjust your graphs so that this is the baseline and then talk about the lack of an "income gap".
If it wasn't on a logarithmic scale, the axis would be so wide as to overshadow the other axises (and there are a lot of axises in those graphs). The absolute values in the range of the "income gap" may not be important when the purchasing power in each locality is taken into account. Like when he mentions the wealth of the UAE only seemed to reduce their childhood mortality rates once education and infrastructure improvements were made. It wouldn't matter if they had even more money (or if the graph was a non-logarithmic scale). His point is that narrowing the income gap isn't some silver bullet that suddenly fixes things.
