Slow productivity growth: lessons from the past

 I am late to this, but this article by the brilliant economic historian Joel Mokyr is fantastic. "(2018), ‘The Past and Future of Innovation: Some Lessons from Economic History’, Explorations in Economic History, 69, 13–26". https://doi.org/10.1016/j.eeh.2018.03.003 

My notes.

1. What are the lessons from economic history to understand current slow growth?
2. It might be that slow TFP growth is the wrong metric for economic transformation.  "To put it differently, students of contemporary technological progress should wean themselves of TFP-fetishism; aggregate measures such as GDP (the basis for TFP calculations) were designed for a wheat-and-steel economy, not for an information and mass-customization economy in which the service economy accounts for 70–80% of value added." 
3. Three reasons why economic growth was so slow before the industrial revolution
1. Malthusian dynamics
2. Smithian pre-1750 economic growth: "growth based on gains from trade and factor mobility, better-functioning and more integrated markets, and improved allocations due to better institutions. Most of the rise of richer regions and towns in medieval and early modern Europe can be attributed to the widening of local trade and the opening of long-distance commerce.  This was vulnerable to rent-seeking: taxes, confiscation, debt reneging"
3. The third factor that explains slow economic growth before the Industrial Revolution is the most obvious and the least discussed one, namely the simple but undeniable fact that people did not know enough about the physical world around them. 
4. Mokyr discusses more of this in the rest of the paper
1. But as I have noted elsewhere, the pre-Industrial Revolution world was limited in its ability to exploit technological advances because even though the pre-1750 world produced, and often produced well. Inventions in the pre-1700 era, however, were normally the result of serendipitous strokes of luck, flashes of brilliant intuition learning by doing, and the slow accumulation of incremental improvements of techniques in use. It was “a world of engineering without mechanics, iron-making without metallurgy, farming without soil science, mining without geology, water-power without hydraulics, dye-making without organic chemistry, and medical practice without microbiology and immunology "
2. of course, not all breakthroughs needed science. Stephenson invented the Rocket with no science training.  So scicence and tinkering were complements. 
3. The breakthrough was 
1. Scientific understanding 
2. measurement
5. what about the incentives to do science?  Essentially it needs a check on rent-seeking
1. "Mercantilism, as Ekelund and Tollison (1981) have pointed out, was a system of rent-seeking, in which resources were redistributed from some groups to others. Mercantilism came under attack from Enlightenment philosophers, and after 1815 it went into a tailspin, replaced by liberalism and a political economy that regarded free trade, open access markets, and a professional and honest civil service as desirable and just."
2. with trade mobiliy and openness "much faster than in the seventeenth century, in the twenty-first century if an idea is generated somewhere , it becomes available everywhere."
6. Measurement. 
1. "If the variety of products and services is introduced on top of the quality improvement, it seems intuitively plausible that mismeasurement has worsened in the past two decades.
2. It also seems likely that the twenty-first century productivity slow-down described by Gordon is temporary, until new Gen- eral Purpose Technologies such artificial intelligence (AI) and genetic editing have fully been incorporated into production lines
3. Moreover, the productivity gains from technological progress in the past two centuries may have been overstated because of inputs that were used and never paid for, in large part because there were no property rights and markets for those inputs. Of those, the physical environment was clearly by far the largest.
4. In many areas, technological progress should thus be seen as a constantly self-correcting process, in which new techniques have unforeseen negative consequences, which require further tweaking, but those fixes in turn will cause more bite-back effects and so on. What negative bite-back effects mean is that the true social costs of many innovations have been understated and the bill for some inputs will be paid by a future generation. If output is generated at time t while the inputs are paid for in t + 1, any comparison of TFP between t and t + 1 will be confounded. TFP growth in the past has been overstated and so its decline may be exaggerated as well, although it is not known by how much. Some portion of innovative effort in the coming decades, rather than aimed at directly raising living standards, may be directed toward maintaining what we already have and correcting the eventual costs incurred as belated bite-back effects kick in."
5. Here is a good example "Such innovations will contribute to economic output and will thus contribute to measured economic growth. But they may not show up necessarily as TFP growth. To see this, consider the following: suppose we have a fossil-burning power plant that produces electricity at, say, 15 ¢per KwH (which is about the US average). Now suppose that we scrap that plant for environmental reasons and replace it with a windmill farm that, with fixed capital amortized in the same way as the fossil plan, can produce electricity at 15.5 ¢per KwH. Using standard calculations (and relying on the dual of TFP computation), this would imply a decline in TFP as conventionally measured of about 3%. But if the windfarm has zero impact on global temperatures, it will have saved on a social cost that is not counted in the standard national accounts."
7. The impact of computers on science has gone much beyond large-scale calculations and standard statistical analysis: a new era of data-science has arrived, in which models are replaced by powerful mega-data-crunching machines, that detect patterns that human minds could not have dreamed up and cannot fathom. Such deep learning models engage in data-mining using artificial neural networks. "
8.