Productivity paradox (aka Solow's paradox)

Computers, Internet, mobile technology, social media, and other digital technologies are so ubiquitous that we cannot imagine our lines without them. Many have even referred to the digital revolution as the third Industrial Revolution. It would seem obvious that all this digital technology has transformed our lives and has led to higher productivity and GDP growth. Yet, the data indicates that productivity growth in the U.S. has not been substantially higher during this digital age. Even as digitization took off in the U.S. starting in the 1980s, productivity growth in the decades since has been lower than productivity growth during the period of 1947-1973. This is known as the productivity paradox or Solow's paradox after Robert Solow remarked back in 1987: “you can see the computer age everywhere except the productivity statistics.” How is this possible? Could it be that this digital age is not as transformative as we think?

I think there are four possible explanations: 

I.  1947-1973 was the anomaly: The period following WWII was unique in American history for several reasons. First, the US economy had suffered through the Great Depression and anemic economic growth for a decade and was likely ripe for a productivity rebound. Second, numerous innovations developed for war efforts later were commercialized and applied in the economy. The unique nature of this period set the stage for enhanced productivity growth for a few decades and makes it incomparable to other periods.

II.    Solow was premature; productivity growth was high during 1995-2005; future gains will take time: The computing hardware advances that began in the 1950s and 1960s and the software advances that began in the 1970s led to the development of database management systems and enterprise software packages. There is some evidence that this wave of digitization caused the significant productivity boost that the US economy witnessed from 1995-2005 (see exhibit).

Just as it took decades for the productivity gains from computing advances to manifest, it will take time for the Internet, mobile technology, big data analytics, and other digital technology to affect the economy. Technology takes time to permeate throughout the economy and society as it becomes more affordable and as use cases are developed. It also takes time for firms to employ the technologies in ways that enhance productivity. This argument is supported by evidence from past GPTs. The mechanical revolution of the First Industrial revolution started in the mid-1700s and productivity gains took several decades to take hold. A similar story can be seen following electrification in the Second Industrial evolution.

III.  The paradox can be explained by the services sector: Over the last few decades, the U.S. economy has been transitioning from a mostly manufacturing-based economy to one dominated by the services sector. This transition could be dampening productivity growth rates (independent of digitization) for two reasons. The first is a composition effect: services generally experience slower growing productivity gains then manufacturing. Education and healthcare are the prime examples of this phenomenon (see Baumoil cost disease) and hence as these sectors comprise a larger share of the economy, productivity gains will mechanically also slow. Second, GDP (and hence productivity) of the services economy are notoriously more difficult to measure compared to the manufacturing sectors. Again, with the rising share of the service economy, the mis-measurement of GDP and productivity could be worsening.

IV.  Digitization exposes the gap between GDP and societal welfare: Imagine a world where robots are not scarce and can provide all the products and services one would desire or need. GDP may in fact be low or even zero but human welfare could be quite high. We are nowhere near that utopian fantasy but even if we have traversed a very small part of the road to that world, one can easily imagine that GDP and social welfare have begun diverging recently and that divergence is leading to a misleading interpretation of low productivity gains. Social media hints at this divergence: access to social media (e.g. Facebook, Twitter, Snapchat) is very cheap or even free and likely below the average consumer’s willingness to pay. Hence, social welfare has risen more than GDP would predict for social media. There are other similar examples in news (e.g. online news, blogs), communication (e.g. email, messaging, peer-to-peer) , entertainment (e.g. fantasy sports), and education (e.g. MOOCs) where the price of the product/service is significantly less than consumers’ willingness to pay.