I was talking to a graduate student who had just taken the written part of her comprehensive exams, and I recalled my own exams a ways ago. Specifically, I mentioned how I flubbed a key question during the oral exam on Comparative Politics: which kind of identity is more or less troublesome for political stability? Shortly afterward, I realized what I should have said: religion > kinship > language > race.
I should have argued that religious differences are most threatening to democratic stability as religions, unlike other identities, imply differences not just in who one is and who one is not (all identities do that), but, more importantly, usually contain different visions about the proper role of government in the society. Religions inherently have political implications by defining what kind of behavior is appropriate, what the right forms of punishment are and who gets to say.
Kinship--where ties to families and extended kin--may be as or more problematic than language now that I think about it some more. If a society is organized around tribe or clan, then any government official will be seen as a partisan, making it very hard for anyone to trust the government and its agents. While suspicion of government can be a healthy thing in moderation, too much undermines the ability to govern.
Linguistic divisions can be problematic as well, as Belgium's recent problems exemplify. Language matters, again not just because it serves to differentiate people, but because how the government regulates languages has direct implications for employment, the educational system, the justice system and the like. Which language that is chosen will influence who can succeed and who is less likely to do so. Changes in language regulations can disenfranchise people, can render people unemployable, etc. While the Sri Lankan conflict has multiple identities in play, including religion (Buddhist extremism might seem like an oxymoron, but this conflict has proven otherwise), but the conflict originated when politicians competed by taking positions over language.
Race, despite its history as the central cleavage in the United States, is not inherently a problematic divide. Two features make it stand out literally: that race is often but not always visible to the naked eye whereas the other key identities do not necessarily have visible markers, and one cannot change one's race despite Vanilla Ice's best efforts (and my campers back in the day when I was a counselor). This properties can make race socially and politically challenging, but different height or skin color by themselves do not lead to political differences.
Over the past twenty years, scholars of ethnic conflict have largely concurred that identities are not static and that several can co-exist. Circumstances may change that lead to one kind of identity gaining more salience than others. And we now see that happening in Somalia.
Somalia has long been divided along clan lines and sub-clans, even before the country fell apart in the early 1990s. I studied Somalia when I was working on my dissertation because it seemed to exemplify the conventional wisdom--that only a country that had no ethnic divisions and risk of separatism could support separatism elsewhere. Given Somalia's inconsisent irredentism, it was worthy of exploration. I quickly realized that Somalia has always faced significant internal divisions and its policies on the kin abroad in Ethiopia, Djibouti and Kenya varied with the relative power of different clan-families in Somalia. Most of the violence before and after the fall of Siyad Barre has been between different kinship groups--sub-clans and clans.
In reaction to the un-ending violence, an Islamic movement gained strength, based on Wahhabism, and has been opposed by the US and its regional ally in Ethiopia. The dynamic of late as been the counter-rise of a Sufist movement, as explained in today's NY Times. We see new alliances emerging across clan divisions, united Sufist against Wahhabi. However, it is not clear who will win and which divide will remain the most salient. Even if the Sufis win, it would be likely that clan ties would re-emerge and undermine whatever stability that might have developed. These problems are not unique to Somalia-they just happen to be more visible and more obviously destructive.