A blog post from BEA Director Vipin Arora
When my kids were younger, they always asked me how things are made. I usually didn’t know. My own dad’s playbook for these questions was saying “I forgot.” I guess I’ve forgotten a lot over the years, too. That said, understanding how things are made is critical to making them better, and ultimately to driving economic growth.
Economists have a proven framework to help with this understanding, arcanely named KLEMS (more on this below). At BEA, we use this framework within the context of our integrated industry-level production account, or ILPA, which is produced jointly with the U.S. Bureau of Labor Statistics. Let me illustrate its usefulness with something that’s on everyone’s mind: AI.
To make and sell artificial intelligence software, companies need servers and storage racks. They also need software developers and data scientists, energy for data centers, materials such as semiconductors, and services such as cloud computing. Put another way, they need: capital, labor, energy, materials, and services—or KLEMS (with a "K" for capital).
The AI vendor knows the cost of each of these inputs. They also know total sales. The KLEMS framework gives us a way to link the two: the value of what’s produced (AI software) and how it’s produced (capital, labor, energy, materials, and services). It also allows us to look beyond one AI vendor to link sales in the entire information industry to those inputs. And if we exclude the intermediate inputs (energy, materials, services), it gives us a way to relate the information industry’s contribution to GDP, known as value added, directly to capital and labor inputs.
This is already valuable—and becomes even more informative when you look beyond one year. In that case, after adjusting for price changes, the KLEMS framework can associate changes in industry value added from year-to-year directly to changes in productivity and capital and labor inputs within that industry. That means if we look across all industries, GDP growth for the entire economy can be traced to growth in industry-level productivity, capital, and labor inputs.
That’s powerful. Let me circle back to AI and show you why.
There is currently no unique industry classified as “AI” in official statistics because production of AI and AI services is spread across sectors. The information industry, however, is one of the primary industries associated with AI development. And our latest update of the ILPAs shows that growth in the capital of the information industry was one of the top contributors to real GDP growth for the entire economy between 2021 and 2024. This is consistent with investment in AI and AI-related services contributing significantly to recent economic growth.
Talk about relevant data! You can’t find that information anywhere else.
As I hope you can see, the ILPAs are an invaluable resource for tracing the sources of economic growth. They can also do a lot more, and I strongly encourage you to give them a try.