The Trivers-Willard Effect (1973) states that optimal female reproductive strategy is to produce sons early in life and daughters later. The assumption is that genes, rather than organisms, are the actual reproductive unit. Traits that help genes make it to the next stage will be encouraged, regardless of costs to their carrier. It follows that a younger, more healthy female will produce sons that are also healthy and more likely - because sons produce more children than daughters - to carry her genes to the next generations. As she ages, her sons are less apt to be healthy. Since nearly all females mate whereas unfit males do not, her next best plan becomes that of producing girls when she is older. The Effect is seen in several species and cultures; the mechanism is unknown.
Estimates are that 1.30-1.62 males are conceived for every female. Differential mortality within the womb gives us ratios wherein there are roughly equal numbers born of each sex. (Spreen O, Risser A, & Edgell D, 1995, Developmental Neuropsychology. NY: Oxford, p 103).
These data have been attributed to antigen production by male fetuses and production of antibodies by the mother. Allergic reactions may well occur between a female and her male fetus (Spreen, et al.). It's also possible that, given less genetic redundancy in male DNA (due to the Y chromosome) males will be able to tolerate a narrower range of embryonic and fetal conditions than females.
Women really are allergic to us and try to kill us early! If valid, then one implication for the Trivers-Willard Effect is that an aging mother may have had prior male fetuses, each one of which heightened her antibody mechanisms. Thus, she may still conceive a greater number of males than females but fewer males survive the pregnancy. If these interactions are analog rather than binary, some sons may be "killed a little bit." This last possibility could account for the pattern that nearly every physical and psychological disability (except depression) affects a higher percentage of males than females. To assess differential survival of late-life male conceptions, it would be interesting to separate male survival data for mothers with a history of producing sons from a those with a history of producing daughters. We might also expect greater male pathology than female with later birth order; such pathology might correlate better with the number of prior males rather than the number of prior children.
If real, this effect could be magnified by an aging mother to the extent that her systems do not operate within the narrower range that is required for the more fragile male children. If times are good, more males will survive. If times are bad, then every egg counts and you need higher parental investment in relatively fewer children; make girls.(1)
1) Wilson (1980, Sociobiology, Cambridge, MA: Harvard Univ. Press) distinguishes r-selection from K-selection. Early in the invasion of a territory having many resources, r-selection patterns usually prevail. More children are produced, there is less parental investment, a higher percentage of instinctive behaviors, and smaller size for individuals. Altruism is more evident. Thus, r-selection appears to overlap somewhat with conditions facing a younger mother ... "produce sons to occupy territory." Later, as a territory fills, K-selection occurs. There is greater parental investment in each child, fewer children are born, individuals tend to be larger in size, and more individualized instruction occurs after birth. Competition between individuals becomes more intense and elaborate. Much of the United States appears to be in a K-selection phase with respect to families, educational practice, and even business organizations.