Timothy Lee explains why classic measurement schemes like GDP fail to register very real increases in value and innovation:

“Official economic indicators are a bad way to measure our generation’s low-hanging fruit.

To understand what makes software-powered innovation distinctive, it helps to contrast it with the industrial-age innovations that proceeded it. For most of the 20th century, innovation was embodied in physical products like cars, televisions, washing machines, and airplanes. This style of innovation is relatively easy for government statisticians to deal with. If an economist at the BLS circa 1961 wanted to know how much the television industry was contributing to GDP, he simply added up the prices of all televisions sold to consumers.

Of course, economists aren’t only interested in measuring national output at a single point of time; they want to measure how the standard of living changes from year to year. If total spending on televisions falls, statisticians need to figure out whether this is because consumers are buying fewer televisions or because televisions are getting more affordable. The distinction is crucial because the former represents a decline in national output, while the latter amounts to an improvement in the standard of living. And of course, economists have to be careful about making apples-to-apples comparisons. For example, the switch from black-and-white to color pushed up average television prices, but it would have been a big mistake to record this as a sign of televisions in general getting more expensive.

There are many important subtleties to measuring changes in economic output, and official statistics have tended to overstate inflation (and hence understate growth rates) to some extent. But the important innovations of the industrial era had some common features that made such problems manageable. They came embodied in discrete physical objects with a fixed feature set. And the value of new innovations was roughly reflected by the prices consumers were willing to pay for them. If consumers were paying twice as much for a 60-inch television as a 40-inch one, it’s reasonable to infer that the former is twice as valuable.

Now imagine an alternate universe in which industrial products did not work this way. Suppose we lived in the world of Harry Potter, and one day in the late 1950s RCA hired a wizard to wave his magic wand and transform all of the world’s black and white sets into color sets. This would clearly represent a large increase in the standard of living—a larger increase, in fact, than the non-magical process whereby people have to buy new, more expensive, televisions. Yet the government in the alternate universe would almost certainly have recorded a smaller increase in GDP. Our own BLS would see consumers buying more expensive televisions while in the Harry Potter universe consumers would be happy with the old, cheap ones. Hence, consumers circa 1970 would be wealthier in that universe than in ours, but official GDP statistics would show just the opposite.

Today these magic wands exist. For example, a couple of years ago, Google waved a magic wand that transformed millions of Android phones into sophisticated navigation devices with turn-by-turn directions. This was functionality that people had previously paid hundreds of dollars for in stand-alone devices. Now it’s just another feature that comes with every Android phone, and the cost of Android phones hasn’t gone up. I haven’t checked, but I bet that this wealth creation was not reflected in GDP statistics. And it’s actually worse than that: as people stop buying stand-alone GPS devices, Google’s innovation will actually show up in the statistics as a reduction in GDP.

Cowen writes that the Internet is producing wealth that “is in our minds and in our laptops and not so much in the revenue-generating sector of the economy.” This isn’t exactly wrong, but it fails to appreciate the extent to which the software industry is entangled with the “revenue-generating sector of the economy.” The digital revolution isn’t just introducing novel ways to amuse ourselves, it’s rapidly displacing a wide variety of “revenue-generating” products and services: typewriters, newspapers, magazines, books, maps, cameras, film development, camcorders, yellow pages, music players, VCRs and DVD players, encyclopedias, landline telephones, television and radio broadcasts, calendars, address books, clocks and watches, calculators, travel agents, travelers checks, and so forth.

Paul Graham and Reihan Salam have been popularizing the term “ephemeralization”, originally coined by Buckminster Fuller, to describe this process whereby special-purpose products are replaced by software running on general-purpose computing devices. As the list above suggests, ephemeralization is affecting a growing fraction of the economy. And with technologies like self-driving cars on the horizon, its importance will only grow in the coming decades.

Ephemeralization offers an alternative explanation for the puzzling growth slowdown of the last decade. Every time the software industry displaces a special purpose device, our standard of living improves but measured GDP falls. If what you care about is government revenue, this point might not matter much—it’s hard to tax something if no one’s paying for it. But the real lesson here may not be that the American economy is stagnating, but rather that the government is bad at measuring improvements in our standard of living that come from the software industry.”