Solow paradox

The economist Robert Solow wrote an often-quoted sentence in 1987: You can see computers everywhere but in the productivity statistics. What he meant was productivity, that is, how much value an hour of work generates in a country. Companies bought computers en masse in the 1970s and 1980s. Nevertheless, this value per hour worked grew more slowly than in the preceding decades. This gap between visible new technology and invisible economic gain has since been called the Solow paradox. A paradox is a finding that at first contradicts common sense.

What the dispute over computer investment decided

Productivity is the most important metric for long-term prosperity. If an hour of work creates more value, wages can rise without prices following suit. If productivity does not grow, the pie to be distributed stays roughly the same size. That is why Solow’s observation was not an academic game but a concrete problem.

For companies and governments, an uncomfortable question arose. Had the billions spent on hardware and software been invested wrongly? Some economists even suspected that computers mainly created work for themselves, such as endless document formatting. Others considered the statistics to be poor, because they do not account for quality and convenience.

In retrospect, there was a resolution. From around 1995 onward, productivity in the US rose markedly for about a decade, and studies attributed a large part of this to computers and networks. The paradox was therefore not proof against the technology. It was an indication that impact takes time.

Why the effect only becomes visible with a delay

The most common explanation is: a new base technology only pays off once the work around it is restructured. A computer on every desk changes little as long as processes, forms, and responsibilities stay the same. Real gains only emerge once a business reorganizes its warehouse, its bookkeeping, and its supply chains. This restructuring takes years and a great deal of money, which first shows up in the statistics as an expense.

A good comparison is the electrification of factories. Early plants merely replaced the steam engine with one large electric motor and kept the old transmission shafts on the ceiling. The real leap came only when each machine got its own small motor. Then the factory floor could be laid out freely according to the workflow instead of the mechanics. Decades passed between the invention and the productivity boost.

On top of this come measurement problems. Much of what software delivers is free or hard to express in prices, such as a search engine or map navigation. Such benefits only partially show up in official figures. Part of the paradox is therefore not an economic problem but a statistical one.

The paradox in the current AI debate

Today the term is mostly encountered in reports about artificial intelligence. Companies are investing enormous sums in data centers and in language models, that is, programs that continue texts and answer questions. At the same time, productivity figures in Europe and the US have so far grown unremarkably. Skeptics call this a new Solow paradox.

In analyses by banks and consultancies, the term serves as a warning to investors. It is meant to remind them that a technology can be real and still not yet be profitable. Conversely, optimists use the same story as an argument: after computers, the upturn simply came with a delay. Both sides invoke the same finding.

A common misconception is to confuse the paradox with a speculative bubble. A bubble means that prices are far above real value. The Solow paradox merely says that a real benefit has not yet shown up in the statistics. For individual companies, AI may already be paying off, while the aggregate economic effect remains small. It is precisely this distinction that makes the term so useful in economic news.