A few weeks back, Robin Harding wrote a nice FT piece on the "productivity puzzle." Yesterday Ryan Avent wrote a really nice note about the notion that productivity growth is very unpredictable. This stuff got me thinking.
Always remember that aggregate concepts are, well, aggregated. Aggregate productivity growth can be usefully divided into (a) productivity growth within businesses and (b) the failure and exit of low-productivity businesses and the creation of high-productivity businesses. Foster, Haltiwanger, and Syverson (2008) found that in manufacturing, about a third of productivity growth comes from establishment entry and exit (the fraction is likely higher in retail).* Another reason to focus on entry is that young firms invest proportionally more in R&D (Acemoglu, et al. 2013). Further, even among incumbent businesses, the effect of firm-level productivity improvements on aggregate productivity depends on the degree to which innovating firms gain resources and non-innovators lose them; the role of reallocation in aggregate productivity growth is therefore huge (Acemoglu, et al. say it's 80 percent).
It turns out that there is lots of productivity heterogeneity among businesses, with Chad Syverson (2011) finding that the 90th percentile (in terms of productivity) is twice as productive as the 10th percentile (in manufacturing, within 4-digit industries).** In some senses, this fact reflects unrealized potential productivity growth. In a frictionless world, unproductive businesses always fail and productive ones always grow. In the real world, the correlations we need are still there, but things may be changing.
In short, productivity growth doesn't only depend on new technology . It also depends on allocation and reallocation of resources--labor, capital etc. If the reallocation machine breaks down, we might not capture all the gains of innovation (depending on the reasons for the breakdown). On the other hand, we can survive a technology growth slowdown if we get better at putting resources where they can be used most productively.
As a side note, recall the Caliendo, et al. paper that found that eliminating shipping costs could boost aggregate productivity by as much as 50 percent. I think there are reasons to be optimistic about our ability to effectively reduce distance: the growth of services, 3D printing technology, technologies that can address the last-mile problem, big data, and so forth. But letting these things work means letting resources be allocated to the firms that are pushing them.
*Note that the data used here track establishments, not firms. I'm being loose with language in this post.
**This paragraph and the one preceding it borrow heavily from literature review contained in some joint work I have with Haltiwanger, Jarmin, and Miranda, forthcoming. And yeah, I have some justified Impostor Syndrome regarding that project.