Tuesday, December 17, 2013

On The Trivers-Willard Hypothesis and Ross Douthat. A look at the Conley&Rauscher Article on Party Affiliation And Having Daughters.#

This post is about the study Ross Douthat was so very happy about:  "The Effect of Daughters on Partisanship and Social Attitudes Toward Women" by Dalton Conley and Emily Rauscher.  The article is behind a pay-wall, sadly, though an earlier working paper, available on the net, gives you a flavor of its arguments.

Rather shockingly, the Conley-Rauscher paper is not about conservative parents cherishing their daughters or traditional morality, as Douthat interprets it.

It is about evolutionary psychology, my dears!  Specifically, the paper argues that the Trivers-Willard hypothesis should explain US party affiliation by parents who have more daughters than sons or vice versa in one data set, published in 1994.  This data set appears to be the only one on US data which has questions about both political affiliation and the sex of the respondent's children.

What is the Trivers-Willard hypothesis (TWH)?   

In evolutionary biology and evolutionary psychology, the Trivers–Willard hypothesis,[1] formally proposed by Robert Trivers and Dan Willard, predicts greater parental investment in males by parents in "good conditions" and greater investment in females by parents in "poor conditions" (relative to parents in good condition). The reasoning for this prediction is as follows: assume that parents have information on the sex of their offspring and can influence their survival differentially. While pressures exist to maintain sex ratios at 50%, evolution will favor local deviations from this if one sex has a likely greater reproductive payoff than is usual.
Trivers and Willard also identified a circumstance in which reproducing individuals might experience deviations from expected offspring reproductive value—namely, varying maternal condition. In polygynous species males may mate with multiple females and low-condition males will achieve fewer or no matings. Parents in relatively good condition would then be under selection for mutations causing production and investment in sons (rather than daughters), because of the increased chance of mating experienced by these good-condition sons. Mating with multiple females conveys a large reproductive benefit, whereas daughters could translate their condition into only smaller benefits. An opposite prediction holds for poor-condition parents—selection will favor production and investment in daughters, so long as daughters are likely to be mated, while sons in poor condition are likely to be out-competed by other males and end up with zero mates (i.e., those sons will be a reproductive dead-end).
The hypothesis was used to explain why, for example, Red Deer mothers would produce more sons when they are in good condition, and more daughters when in poor condition. In polyandrous species where some females mate with multiple males (and others get no matings) and males mate with one/few females (i.e., "sex-role reversed" species), these predictions from the Trivers–Willard hypothesis are reversed: parents in good condition will invest in daughters in order to have a daughter that can out-compete other females to attract multiple males, whereas parents in poor condition will avoid investing in daughters who are likely to get out-competed and will instead invest in sons in order to gain at least some grandchildren.

Bolds are mine.

I'm not sure if the interpretation as given here can be directly applied to human mating habits.  That would assume that human men and women are basically part of a polygynous (one male mating with several females) species, and it seems to ignore the alternative theory that very bad conditions (such as a species almost being wiped out) could trigger an excess of female births and fewer male births simply because that is the way for the species to survive.  Once times are better, a larger share of male births offers the benefits of greater genetic variety.

Whatever the case might be, those who apply the TWH to humans seem to argue that social class, say, can stand in for the kinds of conditions that apply to red deer.  Thus, people who belong to higher social classes are theorized to have more sons and daughters and to treat their sons better than daughters (more food for sons, more education for sons etc.).  People who belong to lower social classes are theorized to have more daughters than sons and to treat their daughters better than their sons (more food for daughters, more education for daughters etc.).  This is not what one observes in general, of course.  And whether, say, the low social classes in a western country are in the kind of "bad condition" that the TWH might apply to can be seriously debated.

Never mind.  Conley and Rauscher grab their version of the hypothesis with both hands.  Thus we get (from the article behind the pay-wall):

One such counterargument is the Trivers-Willard hypothesis (TWH), which suggests that parental investment in sons and daughters depends on parental status (Trivers 1972; Trivers and Willard 1973). Specifically, in order to maximize reproductive fitness (i.e., number of grandchildren), higher-status individuals will devote more resources to sons while lower-status individuals will favor daughters. In other words, parents with low status should favor females, whose reproductive chances are less risky and depend less on external conditions (Hopcroft 2005). In contrast, parents with high status should favor males according to this hypothesis, because their sons will have an advantage in competing for mates and should be able to produce more children than high-status daughters. If these strategic preferences manifest in political preferences, the TWH would expect daughters to produce different political responses depending on parental status. 
Because sons can potentially generate high numbers of grandchildren if they have a competitive edge against other men, they may induce preferences for more libertine social norms and policies—ones where paternal investment is low and restraints on male fecundity are minimal. Meanwhile, daughters may elicit grandparental preferences for a world in which male sexuality is constrained and paternal investment in offspring is greater. 
In summary, in contrast to much previous research we hypothesize that daughters may increase preference for conservative policies in the general population. In the United States, because the Republican Party is generally more conservative than the Democratic Party with respect to family values as well as social and fiscal policies, we predict that daughters will increase parental identification with the Republican Party. 
However, according to the TWH, this relationship should be conditional on parental social status. Among low-status parents (and the general population in contrast to Congress members), more daughters should promote Republican identification. Among high-status parents, daughters should yield no effect unless parents have no sons and expect no further children (which is difficult to measure, but could shift their strategy to favor daughters).

Bolds are mine.

This is not the same as the TWH which has nothing to say about party affiliation.  It is Conley and Rauscher who argue that the Democratic Party platform makes it easier for men to spread their sperm freely and that the  parts of that party's platform which make it easier for women to bring up their children (more support for single mothers, more access to the job market for women etc.) somehow don't cause the hypothesis to reverse.  After all, IF people were actually (unawares) in the business of maximizing the number of grandchildren  then logic suggests that the party which offers more financial support for child-rearing might be the one parents with many daughters would support.

What about the evolutionary psychology the authors use?  It looks to me to be very outdated and simplistic.  For example, later in the article we read:

Males’ optimal reproductive strategy is to sire many offspring with a range of mates and push the parenting requirements onto the mothers.
This is an outdated view of what the potentially optimal reproductive strategy for males might be.  There are several possible alternatives and the idea of spreading-your-sperm-widely is only one of those (an example about spider sex is here).  Because human evolution must have taken place in real (and social) circumstances (and not in the simple model of lots of sperms vs. few eggs), many types of optimal strategies are possible, ranging from monogamous long-term partnering to the scatter-your-speed-while-you-may.

But a more serious problem with all that is the assumption that the TWH so easily translates to party affiliation, and certainly irrespective of the country and the system we are looking at.  For example, the Republican Party in the US is also seen as the law-and-order party, not just the party which is opposed to abortions or premarital sex.  Likewise, the Republican Party favors existing social hierarchies and all which is associated with those, such as the power of men (especially white Christian men) on the top of the hierarchy. 

I could make a seemingly good argument from the Travers-Willard hypothesis which would support the idea that parents with more sons would vote Republican, especially if they are white. I could.  After all, it is that social hierarchy which would let their sons (if successful) have more access to multiple women.  At least according to certain types of evolutionary psychologists.  And I could also make an argument that the Democratic Party is better for the poor parents of daughters, in evolution talk, because it has supported more government assistance to poor people (many of which are women and their children).

So what were the actual findings of the Conley-Rauscher article?  Astonishingly, very muddled, even though they state something different:

Thus, consistent with the TWH, Table IV shows that parental strategic response to the sex mix of children is conditional on parental status. However, the TWH predicted that daughters should promote Republican identification most among low-status parents. We only find this relationship among high-status parents. Similarly, TWH may suggest that having a son first should reduce Republican identification among parents with higher SEI scores and increase it for those whose first child was female. Instead, we find that having a son first makes SEI score positively related to Republican identity, but having a daughter first produces no association between SEI and party identification. 
These results are intriguing and could suggest that having a daughter first generates a more complex relationship between SEI and party identification, given the distinction we discuss above between political party, abortion views, and traditional gender norms, for example. Thus, while the broad pattern is consistent with the TWH prediction that effects are conditional on social status, the specific results are less consistent.
Bolds are mine.

Got it?  The Trivers-Willard hypothesis, in the version Conley and Rauscher support, made them hypothesize that:

However, according to the TWH, this relationship should be conditional on parental social status. Among low-status parents (and the general population in contrast to Congress members), more daughters should promote Republican identification. Among high-status parents, daughters should yield no effect unless parents have no sons and expect no further children (which is difficult to measure, but could shift their strategy to favor daughters).

So lower social status families should invest more in their daughters than in their sons, and because the Republican Party, in the authors' opinion, is the party which maximizes the number of future offspring from daughters, lower social status families should show a greater preference for the Republican than the Democratic Party.

But the study didn't find that.  Rather, it found the reverse.  So I'm not at all convinced that we can say the article supports the Trivers-Willard hypothesis.  In fact, the results of Conley and Rauscher are the opposite of their predictions.

Is there anything else about this paper worth looking at?  The raw data, I think.  As Andrew Gelman wrote, in 2010, the particular data, from 1994 GSS, has data both on political views and the sex of the respondent's children, and the raw data suggests a link between having a girl as one's first child (or as one's second child!) and party affiliation:

So I went into GSS and made a quick crosstab myself, looking at kdsex1 (sex of first child) and partyid (political party affiliation). Here’s what turned up:
- Among the 582 respondents who answered the question and whose first child was a boy: 51% Democrat, 13% Independent/Other, 36% Republican.
- Among the 525 respondents who answered the question and whose first child was a girl: 44% Democrat, 14% Independent/Other, 42% Republican.

(I’ve followed the standard practice in political science of lumping the leaners with the partisans.)
Just for laffs, here’s kidsex2 (sex of second child):
- Among the 435 respondents who answered the question and whose second child was a boy: 51% Democrat, 13% Independent/Other, 36% Republican.
- Among the 429 respondents who answered the question and whose second child was a girl: 41% Democrat, 15% Independent/Other, 44% Republican.

What's going on here, especially given that a recent paper found the very opposite in British data (pdf).

Several possible guesses:  It could be that Americans vote this way, based on the sex of their first child or the sex-mix of their children,  and not the British, for whatever reason (law-and-order comes to mind, in addition to theories Douthat discusses), as a permanent difference between the countries and their political systems.  It could be that Americans in the early 1990s voted that way and might not do so today, because what the parties stand for has changed quite a bit.  It could be that this particular US data set (apparently the only one that has the necessary information) suffers from some kind of sampling fluke.*

Finally to return to our dear friend, Ross Douthat.  He used a study which argues that people are trying to maximize the number of their grandchildren, even if unconsciously, and that this is what might drive parents of daughters more into the strong arms of the Republican Party.  That's not how Douthat read the study, and that is not how he interpreted it for us.

#Apologies for the worst title ever.  I'm tired.  A better one would be How To Maximize The Number of Grandchildren
*Gelman's quote about the correlation between the sex of the second child and the parent's political views suggests to me that there is something weird about the data itself.  Why should we get essentially same findings with the sex of the second child as with the sex of the first child?  Without taking into account the sex of the first child?  Note that out of those whose second child was a girl roughly half already had a boy and so on.