This graph is misleading in two ways. First, note that the vertical axis does not return to zero, but bottoms out at 5 million. Since the height of the graph doubles between 1993 and 1999, viewers get the visual impression that the population has doubled during this period, when in fact it has only increased by 18%.
Frequently, it is necessary to use a graph whose vertical axis does not return to zero, so as to exhibit small but important variations in the data which would be lost on a graph which did return to zero. The usual convention to alert the viewer of the fact that the vertical axis does not return to zero is to put a "wiggle" in the axis; see the next diagram for an example.
It is arguable that graphs whose vertical axes do not return to zero are so common that the sophisticated modern viewer is not misled; the modern viewer doesn't see the current graph as one in which the population doubles, even on a cursory glance. Perhaps we have all become immune to this trick.
The other problem with this graph, however, is rarer, and hence more likely to mislead. The problem is that the horizontal axis changes scale in the middle. From 1993 to 1996 each point on the scale represents six months, but from 1997 to 1999 each point represents one year. This has the effect of making the second half of the graph twice as steep as the first half, even though the population growth is about the same. To the casual viewer, the visual impression is that the population growth rate has suddenly increased.
The source for the population data used to construct this graph is the Census and Statistics Department of the Hong Kong SAR. Note that the graph is my own; the Census and Statistics Department does not commit either of these errors.