Links to Consider, 2/25
Vaclav Smil on the technology slowdown; Nathan Cofnas gives Heterodox Academy an F; John Cochrane and me on the Social Security Trust fund; Emil O.W. Kirkegaard on IQ and political orientation;
Vaclav Smil writes (WSJ),
the overall cost of solar power—measured in dollars per watt of direct current delivered—shows a distinctly declining rate of improvement. Between 2010 and 2015 it fell by 55%, between 2015 and 2020 by just 20%. Even as the costs of renewable electricity generation have been plummeting, the three EU countries with the highest share of energy from wind and solar—Denmark, Ireland and Germany—have the continent’s highest electricity prices.
…the growth of the best processor performance has slowed from 52% a year between 1986 and 2003, to 23% a year between 2003 and 2011, to less than 4% between 2015 and 2018. For computers, as for every other technology before, the period of rapid exponential growth will soon become history.
Seven years later, you can count HxA’s accomplishments in promoting heterodoxy on the fingers of zero hands. It has focused mainly on aggrandizing celebrity academics who hold conventional leftist views, and giving a platform to liberals to engage in empty virtue signaling about their alleged commitment to free inquiry. Scholars whose work is genuinely heterodox have been systematically marginalized.
…Attempts to resist wokeism outside academia have foundered for some combination of the same reasons as HxA. Dissident leaders and institutions (a) are themselves going woke, (b) are squeamish about allying with the Republican Party, (c) make futile attempts to find common ground with woke dogmatists, and (d) avoid confronting the Big Lie about the cause of group differences upon which wokeism is based.
Jonathan Haidt and others, including Jonathan Rauch, are in a no-win position. They are right-coded by the social justice zealots. But genuine conservatives see Haidt and Rauch as wanting to play on the liberal team even though it has been captured by such zealots. It is easy to picture them feeling comfortable showing up at a cocktail party wearing a #neverTrump button. But probably not a button that says #neverDEI.
The ups and downs of the trust fund just reflect a change in how we finance spending. While payroll taxes > social security spending, which was the case until 2007, then payroll taxes are financing other spending. When payroll taxes < social security spending, then income taxes or increases in debt are financing social security spending, which (graph below) was the case after 2008.* The trust fund just adds up this change over time. But exhausting the trust fund is, in this view, really irrelevant.
I think it is best to ignore entirely the (current) financing mechanism for Social Security and Medicare. Instead, think in terms of output and benefits per worker.
Benefits/worker = (retirees/worker)(benefits/retiree)
Retirees/worker are going up because of the Baby Boom bulge, longer life expectancy, declining labor force participation, and slow population growth.
Benefits/retiree are going up because Medicare spending is going up.
Output/worker is going up because of productivity growth.
Let us make this clear with a numerical example. Suppose that today a worker produces 100 bushels of output, and benefits per worker are 15 bushels. After the worker’s output is used to pay benefits, there are 85 bushels per worker left. There is no accounting system that changes that. There is no system of taxation or borrowing that enables the government to turn 100 bushels of output per worker into 15 bushels per worker to retirees and 90 bushels per worker to workers. (Technically, we could trade IOUs to other countries in exchange for bushels of real output, but let’s ignore that here. Also, the bushels will not all be consumed—some of them will be turned into long-term physical assets. For simplicity, ignore that also.)
Now, retirees could decide to give some bushels back to workers via intergenerational transfers. Otherwise, the available bushels per worker are what they produce minus the benefits that are given to retirees. If productivity goes up faster than benefits/retiree, then workers have more bushels. Otherwise, they have less.
Trust fund, schmust fund. It does not matter whether you pay benefits out of payroll taxes, income taxes, or borrowing. There are only so many bushels to go around. When you keep giving more to retirees, the growth in what workers receive becomes low, or possibly even negative.
I have been saying for more than two decades that the “retirement age” (I prefer to call it the age of government dependency) should be indexed to longevity. That policy was never enacted. Perhaps it never will be. Your grandparents vote. Your children and your children’s children do not. All I can say to a young worker is—have a nice day.
So, intelligence correlates positively with wanting more freedom as in social freedoms (abortions, free speech etc.) and economic freedom (less government involvement), but these two political dimensions are negatively correlated. This brings forth the libertarian high IQ rarity pattern. Because the ideologies are negatively correlated, people who are high in both views are rare, but their IQs are particularly elevated. Noah notes that if you combine the political ideologies into a single component, this correlates .40 with IQ.
He uses data sets, such as the GSS, to try to answer the question of whether conservatives really have lower IQ than liberals.
Substacks referenced above:
@
Arnold, Thanks for the simple, incisive explanation of the real economics of social security and medicare benefits.
On Smil’s article: I don’t agree at all. My brother and I were having a conversation last night about a scene in the 1970s Invasion of the Body Snatchers that involves pressing the “button” of a receiver in an old-school telephone and leaving the phone “off the hook” so a person couldn’t call again. This sequence of events completely baffled by teenage niece. If you went back in the 1970s and said “what part of the movie will people be confused about in 2023?” they would never pick out the telephone. But in fact, the “phone” qua phone will be a remarkably short-lived technology in the grand scheme of human technological progress, and what our children and grandchildren will call a “phone” bears only a passing resemblance to its namesake.
And this is my problem with the entire premise of the article: he cherry picks areas that match his conclusions, and proceeds from there. But we will not understand what technologies today will make a significant and lasting impact on tomorrow, and which will vanish. My experience, in my lifetime, is that human technical achievement accelerates at a phenomenal pace. The world that my children, now teenagers, were born into would be foreign to them today. No iPhone, no iPad, no TikTok, no Netflix, no electric car, no mRNA vaccines, no space x reusable space rocket synchronized landings, no video calling, no AI chatbots, etc.
And his semiconductor comment is purely trash, I have to say. That’s the sort of comment you’d expect to see from someone with a New-York-Times-tech-journalist level of understanding of the technology and related industries. Sure, the size of the features you can etch onto a silicon wafer is approaching the point where, you know, there just aren’t many atoms left to play with anymore. But the true innovation, the semiconductor itself and the notion of sequential lithography shrinking as a means to improve performance, was understood half a century ago and the industry has been riding that horse ever since. The fact that a very good (but old) idea is bumping into physics is not a sign of slowing progress or innovation.
To some extent, the semiconductor industry and the industries who benefit from semiconductors have been able to lazily rely, year-after-year, on the steady improvements in power / performance delivered by lithography shrinks. So now what do we do? Well, the very simple answer is make more and bigger chips. The fact that you can only pack so many billions of transistors into a square mm doesn’t say that you can’t add more square mm to get more billions of transistors. It doesn’t say that you can’t put another chip next to the first chip. There are other problems, heat, loss through communication, more complex packaging, etc. that emerge, but those are new opportunities to improve performance. And the incredible amount of money being invested in increases in semiconductor manufacturing capacity around the globe, perhaps for geopolitical reasons but maybe with collateral benefits, should make all of this silicon cheaper than ever before.
You can also change the functionality that you design on a chip to improve performance. Because of lithography shrinkage, the best vector for performance improvement has been the general purpose computing device, programmed through software to execute specific tasks, shrunk as quickly as possible to a smaller transistor size. But that doesn’t have to be the case at all, and we will see more-and-more custom semiconductor development and purpose-built semiconductor designs going forward, because design will become a better vector for improved performance than simple lithography shrinks. The biggest issue here is that students aren’t studying semiconductor design in large enough numbers, but that presumably will evolve.
And he also doesn’t take into account what people *do* with the semiconductor capacity they already have. First off, the centralization of massive computing power into the cloud means that each cycle of compute power is available to be rented out when it’s needed, rather than sitting idly in a corporate datacenter. That on its own represents a tremendous improvement in the efficiency of use of what we already have. Second, what programs are you going to write to run on the computing capacity you already have? It matters a lot. People will use the existing resources to do ever more interesting things, the explosion in AI/ML applications being the most hyped (probably justifiable).
So, in sum, I’ll bet on progress. The question should be, rather, how do we bring greater innovation to areas where there could be tremendous returns for humanity, e.g., energy production, education, human health and well-being, and so on? How do we uncover the lessons of areas where profound progress has been made (e.g., the semiconductor industry, for one) and transfer as much of what we learn from those areas to other areas where progress is needed? That’s the interesting question, rather than bemoaning a supposed “lack of progress” based on some analysis of patents. Please!