Some of the best ideas are wrong

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

Janan Ganesh has a recent Financial Times piece entitled A tribute to great wrong ideas, which argues that flawed models occasionally end up generating useful insights. Ganesh suggests that the Enlightenment concept of the blank slate contributed to the rise of liberalism, the Malthusian Trap contributed to Darwin’s theory of natural selection, and Freudian psychology contributed to innovations in painting and literature.

But the primary focus of Ganesh’s essay is Francis Fukuyama:

The title was too good for the author’s own good. The End of History: it had a definitive ring, and therefore travestied the subtler argument that came in the following hundreds of pages. Francis Fukuyama, who has a memoir out, can be defended on another count. Even if he was wrong to assume the ultimate triumph of freedom circa 1989, he was usefully wrong. His thesis shaped and guided debate for decades. At the margins, it might also have had real-world benefits. The era of liberal hubris that he unwittingly fed probably swept people along who otherwise had illiberal instincts: who might, say, have opposed EU or Nato enlargement. People back the strong horse. He told them which one it was, accurately or not.

The view that Fukuyama was actually sort of correct has now become so popular among intellectuals that it probably no longer counts as contrarian—similar to how the observation that Machiavelli wasn’t all that Machiavellian is now common knowledge. Mocking Fukuyama has become an indicator of having a “mid” intellect.

Here’s Ezra Klein, discussing The End of History:

My view is that very few books in politics are as insightful or as important for understanding where liberal democracy went wrong, why figures like Donald Trump keep rising and why the opposition to them seems so wan and uninspiring so often.This is a book that I think needs to be cleared of a sin it didn’t really commit because it has something to say to us now.

Ironically, Fukuyama did mention Trump in his book:

I actually referred to Donald Trump in my original End of History and the Last Man, when I said that one of the great advantages of living in a capitalist democracy is that it gives you an outlet for megalothymia — for the desire to be recognized as better than other people — in the economy. You could get very, very rich, and that would bleed off some of the ambition and energy that might otherwise go into politics. Little did I realize that, more than thirty years later, that wouldn’t be enough for Donald Trump. So it’s not a guarantee that you won’t have destabilized democratic politics as a result of that kind of outsized ambition.

Matt Yglesias cites another example of a book that is important despite its core thesis being wrong:

For an 80-year-old book written by an Austrian guy in a language he didn’t speak natively, Friedrich von Hayek’s 1944 book “The Road to Serfdom” is incredibly engaging and accessible. It’s also crackling with insights on nearly every page — I kept furiously highlighting passages that reminded me of contemporary controversies and phenomena so I’ll have quotes and points for future articles on everything from antitrust policy to Waymo to health care.What’s remarkable is that even though the book is really good and I strongly recommend it to anyone interested in the future of liberalism in these fraught times, its ostensible core thesis also seems totally wrong.

I’d like to nominate another example. It seems to me that Milton Friedman and Anna Schwartz’s Monetary History of the United States is the greatest book on monetary policy ever written, despite being in an important sense wrong about its core thesis.

Friedman and Schwartz argued that contractionary Federal Reserve policy was largely to blame for the Great Depression. More recently, economic historians including Barry Eichengreen, Peter Temin, Ben Bernanke, David Glasner, and Clark Johnson have suggested that this claim overlooked the key role of the international gold standard. I’d include my own work (The Midas Paradox) within that revisionist group.

So why does Friedman and Schwartz’s Monetary History remain such a great book, if its central thesis is at least partly wrong? I’d point to two reasons:

1. At a time when most people believed the Great Depression showed the inherent instability of capitalism, Friedman and Schwartz showed that the actual problem was contractionary monetary policy. They erred in putting too much weight on the specific role of Fed policy at a time when we were part of an international gold standard, but monetary policy broadly defined was in fact the cause of the Great Depression.
2. The policy implications of their study became vastly more important just 5 years after the book was published (in 1963), as in 1968 we transitioned to a 100% fiat money system where the central bank really does have the ability and responsibility to stabilize nominal spending. In that sense, they were ahead of their time.

Now I’d like to argue that NGDP targeting is another useful but wrong idea. In the past, I’ve argued that while NGDP targeting is not precisely optimal, it is a useful proxy for nominal wage targeting. I’ve also argued that NGDP targeting is likely to be inappropriate in countries where non-wage income is highly volatile, such as major commodity producers like Kuwait. Even so, I suggested that there was little difference between NGDP targeting and a theoretically more optimal approach such as wage targeting for an economy like the US.

So, why did I advocate NGDP level targeting? Perhaps because I couldn’t fit aggregate nominal labor compensation targeting onto my license plate:

It is easier to market NGDP targeting than labor compensation targeting, and I assumed that it made little practical difference. Over the past year, however, NGDP in the US has been distorted by a number of factors, some of which I do not fully understand. Tariffs seemed to cause a big increase in indirect business taxes. And both the tech boom and the energy crisis seem to have sharply boosted corporate profits. This is important because NGDP has four components: wages, capital income, indirect business taxes and depreciation. As a result, NGDP has been rising more than 6.5%/year even as average hourly earnings growth has slowed to 3%:

By early December, 12-month wage growth will likely fall to well below 3%. I did not expect to see this, especially at a time of such rapid growth in NGDP.

But even if NGDP targeting was not precisely optimal in 2026, I’d still argue that NGDP targeting is a useful wrong idea. The actual optimal policy target might be a component of NGDP, such as total labor compensation. Or it might be a Divisia-type index that puts more weight on wage income and less weight on capital income. Either way, NGDP is the place to begin when thinking about stabilizing the economy, as inflation is subject to the “never reason from a price change” problem. NGDP targeting is not perfect, but it’s usefully wrong.

My mistake in this case was not in overlooking the theoretical flaws in NGDP targeting—those are well understood—rather my mistake was in assuming these flaws were not likely to matter for a large, diversified economy such as the US.

Will this mistake be consequential in the future? My hunch is that we’ll go “back to normal” over the next few years, with NGDP once again closely tracking labor income. After that it’s anyone’s guess, as it all depends on progress in AI. If we get the sort of productivity explosion that some are predicting, then capital income might rise sharply as a share of GDP. It’s also possible (but unlikely) that employment would decline sharply. In either case, the Fed should try to maintain a rate of NGDP growth consistent with stable growth in nominal hourly wage rates.

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.