I’ve been writing Construction Physics since September of 2020. Over the past four and a half years I’ve written 186 essays, totalling around 600,000 words. The newsletter was originally focused on understanding the problems of construction productivity (though it’s never been entirely about that), but I have branched out to write more about a variety of topics, including energy, transportation, and scientific and technological progress.
Because the scope of what I write about has gotten so broad, and the backlog has gotten so large, I thought it would be useful to condense a list of some of the major things I’ve learned writing the newsletter. It’s not a summary of every newsletter I’ve written, but a collection of what I think are some of the major high-level takeaways, interesting or surprising facts, and things that gave a major update to my worldview.
Homes, homebuilding, and construction
Roughly 90% of buildings in the US are single family homes, and single family homes make up 60% of the total building square footage in the country.
Prefab construction is typically not any cheaper than conventional, on-site construction, either in the US or around the world. When it is cheaper, it’s often because it takes advantage of low-cost labor, rather than for efficiency reasons.
A big exception to the above is manufactured or mobile homes, which are substantially cheaper than conventionally built homes.
There’s a thesis that the manufactured home industry was irrecoverably wounded by burdensome HUD code requirements in the early 1970s. This thesis is incorrect: the collapse of the industry was much more likely driven by changes in lending standards.
Levittown mass-produced homes were not substantially cheaper to build than other, conventionally-built homes at the time.
Large-scale damage from wildfires is often blamed on extensive use of wood construction in the US, but whether a home survives a wildfire seems to be more affected by things like home age and fire-resistant design details than whether it’s made of wood.
Most attempts at construction innovation have a very long history. Construction startup Katerra’s attempt to achieve low costs through factory-built, mass-produced homes was predated by things like the Lustron Home in the 1940s, Operation Breakthrough in the 1960s, and Toyota’s homebuilding division. Mechanical bricklayers have been attempted repeatedly since the late 19th century.
For as far back as we have data (the late 19th/early 20th century), outside of a few brief windows, construction in the US has never gotten cheaper.
Individual construction tasks have, on average, not gotten cheaper since at least the 1950s.
Bricks haven’t gotten cheaper since the mid-19th century, despite massive improvements in brickmaking technology.
Construction has a reputation for being slow to innovate, but innovations seem to spread in construction at roughly similar rates to other industries, like car manufacturing or agriculture.
Single family homes use less energy per square foot than multifamily apartments, likely because certain "fixed" energy costs like refrigerators and water heaters are spread across more living space.
Historically high US homebuilding rates were in large part driven by falling household size. If you control for this factor, the current low rates of US homebuilding looks less dire.
A common intuition is that people demand variety in their homes, and that this prevents the adoption of mass-production methods for homebuilding, but this intuition is wrong: repetitive designs are very widely used in housing construction.
Construction has gotten much safer over time: not only have death and injury rates in construction fallen, but they’ve fallen much more than in agriculture or manufacturing.
Sears did not pioneer the mail order home. Other companies like Aladdin Homes sold homes through the mail for years before Sears did, and they continued to sell mail-order homes well into the 1980s, decades after Sears left the business.
Construction in general has gotten slower in the US over time, but the contours of these speed changes aren’t uniform, and are sometimes surprising. The speed of New York skyscraper construction has fallen by more than 50% since the 1960s, but in Chicago speeds are down just 10% from the 1960s (though the decline from peak speeds in the 1970s is greater). Today Chicago skyscrapers are built roughly twice as fast as New York skyscrapers.
Even though US skyscraper construction has gotten slower, the US still does well compared to other countries. The US builds skyscrapers at a comparable speed to China, at least for major, high-profile ones.
It's hard to make modern buildings cheaper by cutting unnecessary material out of them: this fruit has mostly been picked.
The modernist, glass box aesthetic took hold in skyscraper construction because it was cheaper to build, and tenants by and large didn't care what the exterior of the building looked like.
Homeowners insurance has steadily gotten more expensive over time. Rising construction costs and more frequent extreme weather events can only explain part of this increase. Increased home size and insurers profits likewise can’t explain it.
Energy
The first windmills for generating electricity appeared in the late 19th century. Wind-generated electricity was common on US farms until the Rural Electrification Administration built out transmission lines to rural areas.
The first gas turbines were built in the early 20th century, but they weren’t economical for large-scale power production until high enough temperatures and compressor efficiencies could be achieved. It wasn’t until after the 1965 Northeast blackout that gas turbines began to be adopted for electricity generation on a large scale.
An important inciting event for wind and solar getting widely deployed and falling down the learning curve was PURPA, the law that required US utilities to buy small amounts of power from independent producers. PURPA was the result of lobbying efforts from a single, small company in New Hampshire that wanted to burn garbage.
After Texas, the state with the most planned electricity generation projects is Oregon.
The US has about 10,000 times as much energy storage capacity for hydrocarbons as it does for grid-scale electricity.
A bottleneck in semiconductor manufacturing is high-purity quartz, most of which comes from a single town in North Carolina: Spruce Pine. Chip manufacturing could probably absorb the costs of switching to non-Spruce Pine quartz, but this would be much harder for solar PV manufacturing.
Three Mile Island didn’t kill the US nuclear power industry: even before the accident new plant orders had ceased, and the industry was dying due to high and unpredictable plant costs.
The US Navy follows many best practices for nuclear reactor construction, such as design re-use, but this hasn’t been sufficient to make nuclear power cheaper than other forms of ship propulsion.
Transportation and other infrastructure
Urban areas have lower rates of power outages, and much lower rates of fire damage and death.
Whenever there’s a major bridge incident in the US we hear stories about the US’s crumbling infrastructure, but the worst bridges in the US are steadily getting fixed. Between 1992 and 2023, the number of US bridges in critical condition declined by more than 70%.
Similarly, US interstate quality is improving over time, and US interstates are roughly as good as major highways in other countries.
San Francisco and Los Angeles have worse roads than any other large US metro.
Tunnel boring speed rose at an exponential rate from the early 19th century through the 1950s, but has slowed down since then. If Elon Musk’s Boring Company achieved its speed goals, it would put us back on track to achieve these historical rates of improvement.
No city grew as fast as Shenzhen in the 20th century, and Chinese cities in general grew faster than anywhere else.
The US has been an uncompetitive, high-cost shipbuilder since the transition from wood to iron and steel ships in the late 19th century. The Jones Act hasn’t helped this situation, but it’s not the cause of it.
AT&T spent substantially more on telephone infrastructure in the 1960s than NASA spent on the Apollo Program.
Modern shipbuilding methods were invented in US shipyards that produced Liberty and Victory ships during WWII. After the war, a US shipyard supervisor, Elmer Hahn, brought the techniques to Japan, who improved and refined them. Several decades later, Japan had conquered the shipbuilding industry, and the US had to re-learn those same techniques from Japanese shipbuilders.
Science and technology
Early on in the development of a technology, a Moore’s Law rate of improvement (doubling in performance every two years) isn’t particularly good. Many technologies improve many times faster than that.
Learning curve improvements, where costs fall by some constant percentage for every cumulative doubling of production volume, are a major driver of technologies getting more affordable, but they can only take you so far. If improvements are difficult to find, costs might stay high and production volumes stay low even at high learning rates. Titanium has a similar learning rate to solar PV, but the latter has fallen in cost much more than the former.
During the second half of the 20th century welding was widely automated by robots, but the jobs that were automated were mostly welding machine operators, not manual welders. In other words, welding robots mostly improved existing automation, rather than automating tasks that had previously been manually done.
Industrial robots of today aren’t much cheaper than industrial robots from the 1970s in terms of payload capacity, but modern robots are on the order of 50 to 100 times more accurate.
More Nobel Prize-winning research has been done at Cambridge than at any other university.
Bell Labs has done more Nobel Prize-winning work than the next six private corporations combined. Most of its prize-winning work was done by researchers hired after the company invented the transistor, at the peak of its prestige.
The prestige of Bell Labs inspired several other companies, like Ford, Exxon and Xerox, to fund similar labs that produced basic research uncoupled from concerns of immediate profitability.
Morris Chang began his semiconductor career completely by chance. He originally planned to take a job at Ford, but got so angry at how he was treated when he asked for a higher salary that he took a job at electronics manufacturer Sylvania out of spite.
Rocket scientists weren’t the only Germany assets that the US and Russia divided up after WWII. They also each acquired some of Germany’s very large forging presses.
Research and development programs have generally gotten more expensive over time, but some things have always been pricey. Development costs for a new jet engine, or for a new commercial airliner, have been in the billions of inflation-adjusted dollars since the 1950s and 60s.
Nearly every aircraft the US produced in large numbers during WWII started design work before June 1940, more than a year before Japan attacked Pearl Harbor.
Making a microchip requires thousands of process steps, and a silicon wafer might spend months going through the entire process.
My main takeaway from this list of takeaways, and from writing the newsletter more generally, is that there’s almost always more to the story. Things that seem like recent developments often have key predecessors going back decades or even centuries. What seem like historical inevitabilities are often highly contingent products of chance and circumstance. Intuitive explanations for phenomena are often wrong, and even when they’re right, the full story is often much deeper and more interesting. Causes are often complex, and what seem like simple problems often tenaciously resist solutions.
This complexity is what makes writing Construction Physics so fulfilling to me. Here’s to the next 600,000 words of it.
I really enjoy reading your posts.
What a great factoid summary of my favorite newsletter! Not only is reading Construction Physics absolutely fascinating and supremely edifying, it gives me hope that there's still a future where rationality and a shared objective reality still exists as the basis for humanity to progress forward and avoid falling into an ignorant abyss.