Several commentators have mentioned that socioeconomic/intelligence factors may explain the workings of this graph.
Whilst I did not think socioeconomic factors would alter things much, I could not prove it. Fortunately the source of this information has been found, (Thanks Leah!) and what's below is taken from it.
As you can see, socioeconomic factors don't play any major role in the number count. Furthermore, intelligence--measured imperfectly by its proxy education--seems to have a small effect. Highly educated and affluent white girls seem to be doing their fair share of the heavy lifting. Must be those Law/arts majors.
The paper from which this data is from can be found here. There is literally loads and loads of information in it.
The 2002 data is here and pretty much confirms the 1995 findings. Table 11, page 29 for those interested.
With regard to the 1995 data, there is a flaw in the Heritage Foundation's methodology. Apparently Married= Married+ remarried. Approximately 9.3% of the Married group were remarriages. I suppose that's why the Heritage Foundation labeled the graph, Stable Marriages instead of First Marriages. Anyway, what this does is artificially elevates that stable marriage rate of the 2 or more partner group, since about 15% of the marriages are remarriages.
It also dilutes the 1 partner group falsely by about 3% making the probability of stable marriage lower than what it is in reality.
I've got to admit that whilst I thought promiscuous women were less safe bets, I never imagined that the effect would kick in after so few partners.
It's actually spurred me to do a bit of a non systemic literature search and the data which I have found by other independent researchers seems to point in the same direction. I'll hopefully put up a post in the next few weeks when time avails.
This has all been rather profoundly depressing. Based upon the data, the current "hook up" culture is likely to produce a social disaster never before seen in the West. We are facing a social apocalypse. Society is going to implode.