Growth in the US and Europe has been falling for several decades – from 4% per year in the 1950s and 1960s to 2% in the 2010s. The major driver of slowing growth is the slowing rate of productivity growth (Fernald 2015). And slow growth is a major factor behind political protest movements that contributed to the votes for Brexit and President Trump. Will this slowdown in growth persist, or is a ‘fourth industrial revolution’ about to begin, possibly encouraged by breakthroughs in computing, robotics, and artificial intelligence?
This debate has morphed into an argument about how many innovations are out there to discover. Optimists and pessimists alike think of discovering productivity improvements as similar to drilling for oil. There are quantities beneath the earth’s surface waiting to be pumped out – but there’s a fixed amount, and once it’s gone, it’s gone. Pessimists like Gordon (2016) argue that we’ve extracted almost all the ‘oil’ of productivity growth already. Nothing as life-enhancing as the indoor toilet could possibly still be out there awaiting discovery. Optimists like Brynjolfsson and McAfee (2014) suggest that new technologies signal endless golden geese, ready for plucking.
We argue that a single-minded focus on the quantity of undiscovered ideas is unhelpful. It is not just how many ideas for productivity growth are left, but what it would cost to get them out of the ground – and, crucially, how much we’re prepared to spend to do it. For a long time, geologists have been forecasting ‘peak oil’, only to be surprised by new deep-sea discoveries and shale oil. We, likewise, see a continuing stream of innovations. But, just as newer oil sources are increasingly costly to extract, coming up with new ideas is getting more expensive.
Measuring the cost of ideas
In a new study, we show that the costs of extracting ideas have increased sharply over time (Bloom et al. 2017). In other words, the innovation bang for the R&D buck (or ‘research productivity’) has declined. In an accounting sense, therefore, low productivity growth in the economy is a direct consequence of research effort failing to increase fast enough to offset declining research productivity. If we want to restore economic growth, we need to pay for it.
We measure declining research productivity in a wide variety of industries, products, and firms. To do this, we find places where we could measure both the inputs to research (research expenditure or effort) and the outputs from the same research (productivity improvements) over time. Wherever we look, the amount of research effort needed to achieve a given proportional improvement in productivity has increased – often dramatically so.
Consider the aggregate level. Figure 1 plots two lines. The first, in blue and corresponding to the left vertical scale, is US total factor productivity growth (the part of output growth left after accounting for inputs of labour, capital and other factors). It declines gently over time. This is our measure of the output of economy-wide research.