Unfortunately I've forgotten a very famous social psychology experiment wherein one group (group A) was allowed to dictate their preferred wage difference between their group and and another group (group B). They chose the option which gave them the least in an absolute sense because the option gave them more than group B by comparison. They were divided according to profession. It's a very famous experiment, so I'm sure someone here will know it.

In Irrationality, Sutherland cites Brown (1978, "Divided we fall: An analysis of relations between sections of a factory workforce") and states:

In real life, the rivalry between groups may be so irrational that each may try to do the other down even at its own expense. In an aircraft factory in Britain the toolroom workers received a weekly wage very slightly higher than that of the production workers. In wage negotiations the toolroom shop stewards tried to preserve this differential, even when by so doing they would receive a smaller wage themselves. They preferred a settlement that gave them £67.30 a week and the production workers a pound less, to one that gave them an extra two pounds (£69.30) but gave the production workers more (£70.30).

In a highly-cited review, Tajfel (1982) states:

An intriguing aspect of the early data on minimal categorization was the importance of the strategy maximizing the difference between the awards made to the ingroup and the outgroup even at the cost of giving thereby less to members of the ingroup. This finding was replicated in a field study (Brown 1978) in which shop stewards representing different trades unions in a large factory filled distri­bution matrices which specified their preferred structure of comparative wages for members of the unions involved. It was not, however, replicated in another field study in Britain (Bourhis & Hill 1982) in which similar matrices were completed by polytechnic and university teachers.

A brief look at recent studies seems to suggest a more nuanced relation, but I'm not familiar with the literature. See, e.g., Card et al. (2010).