Conditioning on this, and all else being equal (assume prior of zero information / even spread of probability distribution), I simply have to take the 40$k. On a zero-information prior, the probabilities dictate that my intellectual time is most likely more important than that of a random other person, and that it is most likely that my job is not an optimal-utility one, such that I can do more good or produce more utility in my spare time than my spouse could in the same spare time obtained with the lower-paying job.

You only need to have an IQ / other measure of social-utility-production above the average (or mean? I haven't actually worked out the maths, just replaying this from rote-learned stuff) in order for this to hold true, in terms of expected utility when you know nothing a priori about the spouse.