Innovation Doesn’t Equal Productivity, and Patents Don’t Always Represent Innovation

The prevailing view in economics is that innovation drives productivity. And that patents represent innovation. And that R&D drives patenting.

It’s as if the model is bivariate, where Ix = Px (where I is innovation, P is productivity, and x is the amount of both).

And because R&D and patenting have increased in the last 15 years, this raises a conundrum: Why has this not translated into higher productivity, which has been anemic over the previous 15 years? To paraphrase Robert Solow, we see innovation everywhere except in the productivity statistics.

Economists’ reliance on R&D and patent metrics distorts our understanding of productivity growth. This matters because it leads policymakers and others to believe that they can solve the Western productivity crisis with more innovation.

That assumption is simplistic at best and wrong at worst.

It’s time to correct the conclusion. There are three reasons why these proxies fail to capture the forces that actually drive productivity:

• First, lots of innovations have little or nothing to do with productivity.
• Second, lots of productivity gains have little or nothing to do with innovation.
• And third, not all productivity-related innovations are created equal.

To see why the first reason is true, I examined a random sample of 15 patents issued by the United States Patent and Trademark Office (USPTO) for the week of October 28, 2025.

One through ten on the list below appear to have little to no productivity benefits—a new kind of smartphone grip or a better hookah is not exactly ushering in a major productivity surge.

Eleven through fifteen appear to have some potential productivity benefits, but they seem incremental at best, such as an improvement to a grounds-maintenance vehicle or an apparatus for vertical farming.

Below is the full list of randomly selected patents:

1. Metatarsal Arthroplasty Devices
2. Animal Monitoring and Deterrent Device
3. Paint Brush Handle With Integrated Hanger Assembly
4. Method of Spraying Small Objects
5. Grip for a Foldable Electronic Device
6. Nanoemulsions Comprising Fatty Acid and N-Acyl Derivatives of Amino Acid Salt
7. Disposable Weather-Protective Animal Garment
8. Hookahs, Heating Units, and Related Methods
9. Wearable Leg Warmer
10. Golf Shoe With Spring Plate
11. Apparatus, System and Methods for Improved Vertical Farming
12. Bulk Shrub Harvester and Related Methods
13. Plants and Seeds of Hybrid Corn Variety Ch010374
14. Self-Cleaning System and Method for Extraction Cleaners
15. Implement and Grounds Maintenance Vehicle Including Same

At the same time, lots of things that boost productivity are not based on new innovations.

Redesigning cars and parts, as the Chinese do, to reduce the number of components in a vehicle—and thereby boosting assembly productivity—is not based on R&D or patents, but rather on more careful industrial design and a willingness to upset the status quo that says a particular car has to have a particular number of parts.

Many organizations still have not moved to cloud computing, even though it’s more efficient and cheaper. Yes, cloud was an innovation, but getting the full productivity benefits requires widespread adoption.

Telemedicine for routine doctor’s appointments can save time, but it is still underutilized, in part because of insurance rules.

Productivity is also dragged down by small businesses. Yet public policy continues to favor small firms in ways that enable them to hold a larger market share than their productivity would otherwise warrant.

There is a final big reason for the gap between innovation amounts and productivity rates. For over 75 years, economics has been based on statistical analysis, and innovation statistics have a very hard time differentiating between what Clay Christensen called disruptive innovation and sustaining innovation, or between general-purpose technology and incremental technology.

Economists look at all 15 patents listed above as 15 innovations, full stop.

But what really drives productivity is the emergence of critical general-purpose technologies that substantially increase in performance, decrease in price, and are used in a wide array of functions across the economy.

As I wrote in The Past and Future of America’s Economy: Long Waves of Innovation That Power Cycles of Growth, steel was a GPT, as was electricity, electronics, the postwar chemistry revolution, and, more recently, computers and the internet. But these true, initial GPT breakthroughs are rare.

When these GPT technologies move up the S-curve and become cheap enough and powerful enough, they are widely adopted and drive productivity growth for two to three decades. The problem now is that we are between technology long waves.

We have no technology driving the next big surge in productivity. Certainly not a bulk shrub harvester. Not the smartphone. Not crypto. Not 5G. Hopefully, AI and the related technologies of autonomy and robotics will be that next big GPT. But they certainly aren’t now.

LLMs like ChatGPT and Claude, for all their hype, are extensions of past internet technologies, not fundamental changes. And robotics are not cheap enough or good enough to perform most physical functions. They are a bit like electric motors in the early 1900s: expensive, not very versatile, and underpowered.

It took another 10 to 20 years of technological improvement before electric motors finally began to appear in productivity statistics. Hopefully, we won’t have to wait that long for the next GPT to mature.

And this gets to the key problem in economics: As a statistical field, it is largely useless in explaining the slowdown of productivity growth or how it might turn around. For that, one needs a deep, qualitative understanding of technology, science, and engineering—not patent counts.

If this is the case, then much of what economists look at to understand growth is largely meaningless. We could have the best regulatory system, the best tax system, etc., and without a strong GPT, the natural limits to growth will remain relatively low.

So, it’s unrealistic to expect economists to develop a national GPT strategy. But others, those with a deeper understanding of science, engineering, and business, can contribute to that. And that is what we need.

Imagine if government policy could accelerate the maturity of the next GPT by a decade and then accelerate adoption on the S-curve of full deployment by another decade. Now you’re talkin’ real money.