These crude rates are misleading at best. To get an accurate picture you have to break it down into a large multitude of demographics:
- Educational attainment
- Ethnicity (including whether both are the same ethnicity, and if not, which of the man and woman is what ethnicity; for example in America black man+white woman like my parents = higher risk of divorce than white+white, but white man+black woman = lower risk of divorce than white+white, and asians overall have the lowest risk of divorce)
- Income
- Age of marriage
- Age of first sexual activity (could only find this for women so far)
- Number of previous sexual partners (again only found a study that shows correlation for women so far)
- Religious attendance (this is also the one factor that lowers odds of
cheating among both genders)
- Previous divorces (you did note this)
- Occupation (the divorce rates actually vary wildly by occupation)
The odds of divorce are
completely different depending on which demographics the couple fits into, to the point that it's practically irrelevant to group them all together into into one statistic and use it as a bludgeon.
I haven't yet found a comprehensive regression analysis to begin sorting out which variables are more symptomatic of other variables and which are more causative (for example, the fact that higher education is associated with lower divorce may be because high education leads to higher income which is what really matters - or they may both matter, or other factors that lead to divorce may just be more concentrated in people with lower education), but I'm looking for it. Without a comprehensive regression analysis, even all of these demographic factors are just guesses in the dark too statistically speaking.