What I have been reading: November edition
Women are more interested in relationships, and men are more interested in ideas and the physical realm: literature edition. Also see my recent blog on the great feminisation thesis for more on sex differences; without the misinformation nor any ulterior agendas.
Log-transformations are standard in economics (and in general most emprical social-science research), yet how should you handle zeroes?
An underrated case for markets: talented individuals yield incentives to innovate and coordinate production, which produces positive spillovers. In alternative systems, they are likely to become rent-seekers. How we allocate talent matters for endogenous growth.
Was most of the post-war baby boom due to declining maternal mortality? If so, then most of the prevailing commentary explaining declining fertility (housing costs, shifting gender roles, etc) are downgraded in their explanatory power.
On the importance of interviews for job search, for candidates.
Ayn Rand on deregulation. I agree: libertarians cannot ignore state capacity. What I dislike most about the culture of my ideological group is the tendency to ignore the details of policy implementation. Advocating for a nightwatchman is all well and good: how will it look in practice? What will be the alternatives to domains where the state currently dominates the supply-side (e.g. redistribution)?
In his latest post, Scott Alexander highlighted the tendency for LLMs to interpret language literally whilst neglecting pragmatics or contextual cues. I ask ChatGPT whether this makes AI autistic? If so, then we have a bullish case for the cognitive strengths of autism.
Lucas Critique: biodiversity edition. Let's not use linear extrapolation from past aggregates to predict the future of the environment. Technological progress works in “solving” climate change.
Contrary to intuition: ambiguity and uncertainty in court cases is actually a signal of guilt.
Roger Farmer argues that, to increase predictive accuracy, you impose ad-hoc additions to the NK model. Yet this inevitably trades off the very internal consistency that is the rationale for NK DSGE in the first place. My approach to this dilemma, as fits my general epistemological reasoning, is Bayesian (not an intentional pun on Smets and Wouters!): use NK DSGE as your prior or baseline theory, then update based on which assumption you need to relax for your targeted mechanism. Economists should learn from superforecasters here!
Yes. I'm with Friedman in my approach to economic modelling.
Most surprising finding ever! At the risk of losing some popularity, it's indeed true that misinformation is primarily a problem on the right. Education polarisation, and the drive towards populism (I define such here), will inevitably make the right more stupid. Yes, you could draw some equivalence with (say) gender ideology or climate hysteria on the left, yet a flawed model of the world is nowhere near the same as outright falsehoods.
Strong publicised political stances are bad for business. Whether or not you are willing to tradeoff some monetary success for less self-censorship (which I do to some extent) is another matter altogether however.
Matt Yglesias on the great feminisation. Yes, to the extent that it reflects increased specialisation and division of labour, of course second-wave feminism is a good thing. So yeah, perhaps I’ve changed my mind, and am no longer anti-feminist. Perhaps I even subscribe to some variant of second-wave feminism (one grounded in respect for the individual regardless of background, not in seeking equality of group outcomes). It's a good thing to change your mind, and we need more of it.
The first thing I learnt in undergrad was that controlling for Z does not mean causality. Again, let's teach statistical literacy in high school as a mandatory subject!
Behavioural economics replicates. Indeed, lab experiments simulating markets (for instance, the works of John List and Alvin Roth) are some of the discipline’s best work. Of course, external validity is an issue, but one easily resolved via Kahneman’s and Tversky’s distinctions between S1 vs S2 thinking. My priors: decisions with longer time horizons (like investments or durable goods purchases) involve more S2 relative to S1 thinking, whilst good market design and incentives can prompt S2 over S1. By the way, what are the implications for macroeconomics if agents are not rational, so cannot form rational expectations? Rational expectations allow inflation to fall rapidly post shock without a Taylor rule, which is exactly what happened post-COVID. Strictly speaking, REH is false, yet leads to much more accurate predictions than alternatives, such as adaptive expectations. Another case for Friedman's instrumentalism…
The 1980s anti-alcohol campaign in the USSR reduced homicides, suicides, and unintended fatalities. More surprisingly, fertility rates also rose, in tandem with divorce rates. Is Tyler Cowen right on alcohol? Increasingly, I'm included towards that view, if indeed we care about pronatalism. With tentative (albeit imperfect) success, I'm adopting teetotalism myself. A subject for a future blog post…
Fellow Brits and Europeans: invest in air conditioning!
The gender divide: politics edition. Interestingly, more Gen Z men than not consider toxic masculinity an issue in society. Gen Z males may reject the blank-slate, but that certainly does not imply an endorsement of bigotry.
Rising markups do not necessarily imply reduced dynamism. Covariance in market share and firm-level productivity is likely to be the elephant in the room here.
Rubies have recently been observed by the JWST in our early universe. “Little red dots (LRDs) are a hybrid between a black hole and a star: an active black hole wrapped in a cocoon of hot, dense gas — much like the atmosphere of a star — that glows as the black hole warms it.”
A pushback to the nascent claim that ketamine could treat depression.
On AI and free-speech. This maps well with the general state of freedom of expression across countries. Unsurprisingly, Grok scored the highest for "free-speech culture".
Indeed. I see tort law as a return to common law traditions, which have increasingly been eroded via state overreach in statutory regulation.
The poor disproportionately make bad decisions, which helps to explain their predicament. Counterintuitively, we should be glad that this is the case: a) provided one is prudent, “bad luck” shocks are transitory and dissipate over time; b) we don't need to do the impossible and change the world to help ourselves, which implies one should be hopeful and optimistic.
Embryonic selection in IQ. This utterly refutes the anti-hereditarians. If heritability of traits was zero, then this would not be possible. Yes, the different methodologies to estimate heritability are imperfect, yet they are good enough to boost IQ by almost 10 points.
Although for some individuals there is a biological basis for their gender dysphoria, for the overwhelming majority, transgenderism is a cultural fad. Kauffman et al are indeed correct here.
For an authoritarian government imposing sweeping restrictions on phone use and video gaming, the effect sizes are very small (and not compared with the benefits of mobile phones: you would not be reading this informative blog in the absence of social media!). We should be sceptical of the claims of Haidt et al.
“Slavery made the South poorer because it kept the South rural -- it prevented the urbanization and agglomeration” that drives endogenous growth.
Very high birth rates amongst the Amish. I note too that Israel is the only developed country with above replacement-level fertility. The Haredim and Hasidic communities are known for their fertility. What can the rest of us learn from these cultures?
Yes, the discourse has become a lot angrier in recent times. Anger also fosters zero-sum, in vs out group thinking. Anger is probably the worst emotion of them all in terms of the flourishing of humanity. The market for hitmen speaks for itself on this!
Given that different LLMs play different strategies in these games, do they have different personalities? If we can attribute a personality to an LLM, is there a sense in which they display consciousness?
Cancel culture and threats to free-speech: neither a left nor right-wing phenomenon. No one really likes freedom of speech for their opponents, which is exactly the case for why we must strenuously defend it.
Hey, great read as always. I especially loved your thought on LLMs interpreting language literally and asking if that makes AI autistic. The "bullish case for cognitive strengths" line was brilliant; definitely resonated with my own fascinaton for AI and its quirks. So insightful!