Things to think about #4

Mothers and money; a birth penalty still?

The Economist’s Free Exchange column drops in on the question of an economic motherhood penalty from childbirth. It is nice to see that the Economist correctly distinguishes between two distinct economic motherhood penalties, both of which can be traced to the interplay between evolutionary forces and modernity, where the latter in this case is defined as an environment with rapidly increasing returns to investment in human capital and education. The first, between fathers and mothers, emerge because the cost of child-rearing especially in the early part of a child’s life overwhelmingly falls on the mother, a conclusion which follows from Trivers (1972). This is true in terms of the cost during pregnancy and immediately after too. It is also true before we consider the possibility that the resource allocation trade-off for many women shifts in the wake of motherhood. The second motherhood penalty occurs between women. Put simply, in an economic structure where childless women have the ability to devote all their resources to somatic investment and take advantage of the above-mentioned increasing returns to human capital investment, the wage and wealth divergence between women who have many children and those who have none will widen significantly, at least in theory. For more on this, I cover the theory in more detail in my essay on fertility and sexual selection; see here.

Evolutionary hypotheses are one thing, but what about practice? Is the motherhood penalty a thing? The literature suggests that it is. Kleven et al (2019), cited by the Economist, finds strong evidence of a motherhood penalty in developed economies, and Cukrowska-Torzewskaa & Matysia, (2020), a meta-study, finds a motherhood wage penalty—between women—of around 4% driven by the loss of human capital across women’s reproductive careers and active employment choices after the birth of the first child.

The new evidence presented by the Economist cites studies investigating earnings post IVF treatment to isolate the effect of women’s earnings for births which are planned and wanted. This new evidence suggests that the motherhood penalty falls over time. From the first study, using data from Norway, the Economist summarizes the results as follows;

The Norwegian study followed women undergoing ivf treatment for roughly a decade, finding that the annual incomes of those who gave birth fell by an average of 22% in the short term, compared with those who remained childless. In the long run, however, this penalty narrowed to just 3%. Fathers’ incomes did not drop; instead, they rose by around 10% over the long run.

A similar Danish study finds an even more encouraging result, extending the period of study to a quarter of a decade;

A new Danish study considers a longer timeframe. Petter Lundborg of Lund University, along with co-authors, looked at up to 25 years’ worth of data from the point of each woman’s first ivf treatment. They found a similar pattern: a sharp short-term drop in mothers’ earnings, but no long-run earnings penalty compared with women who did not have children. In fact, their data reveal a small “motherhood premium” after about 15 years, which over a lifetime more than compensates for the initial drop.

So, does the motherhood wage penalty evaporate over time? Perhaps, but there are at least two potential sources of sample bias in the studies above, both highlighted by the Economist.

First, while using IVF treatments is a good way to isolate the effect on the income of mothers from planned births, it risks biasing the sample towards women with relatively high education, and crucially, high income and wealth since IVF treatment can be very expensive. I am not sure in this case whether the studies noted above only considered IVF treatments for women who were subsidised by governments to undergo treatment.

Second, Denmark and Norway are countries with very well-developed and generous publicly funded child-care and -support programs. This could be the driver of a falling motherhood wage penalty over time in these studies, rather than some generic long-term tendency of this effect to fade over time.

Thirdly, and more importantly, why correct for the randomness of births altogether? Indeed, it seems to me, a priori, logical that any motherhood penalty of planned births should be less than from random or unplanned births, and I don’t see why we would want lose sight of that. If we admit that some births are unplanned or even undesired—despite the significant ability to day to control births—the reverse methodology from the point of view of the studies cited by the Economist is to analyse births as random income shocks, from which mothers suffer the largest cost in terms of income and wealth. This at least should be part of the overall discussion. Similarly, when women have their first child also would seem, based on the results above, to have a potentially significant impact on any motherhood wage penalty.

Finally, the elephant in room in this discussion is whether social structures and policy can be used to mediate any inherent motherhood wage penalty? Certainly, the evidence of a relatively smaller motherhood wage penalty in Scandinavia suggests that policy has a role to play, but can the penalty be eliminated completely, between genders and between mothers and childless women? I have my doubts, but the work and study will go on.

Are we Losing ourselves to AI?

One conclusion from Matthew B. Crawford’s recent Hedgehog Review essay on AI—which originally appeared on this Substack—is that a father shouldn’t use ChatGPT to write a speech for his daughter’s wedding. But there is a bit more to it than that. Matthew starts out on a sombre note;

Elevated “deaths of despair” and declining birth rates in the West must be due to an array of factors, hard to tease apart. My hunch is that one of them is what the sociologist Richard Sennett called “the specter of uselessness.” He meant feeling redundant at work. But there is a deeper, existential version of this that may arise when the world feels already-occupied, so there is no place for you to grow into and make your own.

This rings true to me, and AI likely is feeding such a spectre of uselessness in ways that we haven’t yet realised, let alone articulated. Michael starts his inquiry by noting that the ability of large-language-models (LLMs) to generate language and text with speed, and in vast quantity, risks robbing humans of a fundamental part of being and existing in a modern cultural setting. If people don’t actively engage in the creation of language, and if people aren’t putting themselves on the line in doing so, they don’t just lose agency, they lose a fundamental part of being human in a cultural world. According to Michael then, outsourcing communication to an LLM AI robs people of significance in a world where many already feel disenfranchised in so many aspects of the economic and political culture.

“We care” Michael says, which is because

“… unlike an LLM or a parrot, things have significance for us, and we search for words that will do justice to this significance.”

Proponents of AI will undoubtedly jump in here and note that over time, and with practice, the AI will be able to emulate any such significance and care with almost perfect fidelity. That’s probably true, but this isn’t Michael’s point, or perhaps this is exactly the point? In a culturally complex and rich world, the process of creating language matters for people just as much as the end product, if not more. After all, the father who pens his own speech for his daughter’s wedding might not, based on some measure of objective quality, reach the oratory heights of a Barack Obama, Martin Luther King or Christopher Hitchens. But it will be his speech, and that matters too, more even. Michael gets to the nub of the issue when he says;

What would it mean, then, to outsource a wedding toast? To use Heidegger’s language, some entity has “leaped in” on my behalf and disburdened me of the task of being human. For Heidegger, this entity is “das Man,” an anonymized other that stands in for me, very much like Kierkegaard’s “the Public.” It is a generalized consciousness—think of it as the geist of large language models.

The idea that LLMs have the capacity to temporarily displace, or disburden, what it means to be human is a profound one, for at least two reasons, I think. First, it speaks to the allure of AI in the first place. After all, who wouldn’t want to disburden themselves of more difficult parts of the human conditions from time to time, especially if it can be done, in many settings, without cost and without being found out. Secondly, do we have here an explanation for why AI LLMs has come for the creative arts much more aggressively, and successfully, than expected? As the by now famous meme from sci-fi author Joanna Maciejewska puts it;

You know what the biggest problem with pushing all-things-AI is? Wrong direction. I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes.

The explanation for this counterintuitive volte face in the dynamics and proliferation of AI, contrary to expectations, is not just that it be can nice, once in a while, to disburden oneself for the hard work of being a cultural human. But in an economic sense, it is also where the relative bang for the buck is, at least in some areas. After all, creativity is hard and comes with inherent risks of getting it wrong. Why shouldn’t we attempt to boil down the parameters of such a process and feed it to an AI? Well, as Michael says at the finishing line, after lamenting that AI LLMs will tend to cluster content, interaction and language at a boring mean:

This mood of interchangeability is likely to deepen as AI saturates the world and we are tempted to let it stand in for our own subjectivity. But, like that father at his daughter’s wedding, we are still free to refuse it.

We are indeed.