There’s more innovation than ever – but instead of technological breakthroughs, we just get 5000 different breakfast cereals

Lesen Sie die deutsche Version hier.

Research and Development (R&D) spending by firms is widely seen as an engine of productivity growth, increasing economic output more than inputs. Thus, it is encouraging that businesses have increased R&D investment sharply. US business R&D has grown more than 20-fold since 1980 and has more than doubled relative to GDP. Similar patterns are seen in other developed nations. But researchers have noted that despite this increase, productivity growth has been stagnant or declining.

Some economists such as Robert Gordon, focusing on information technology, contend that R&D today just isn’t effective at bringing productivity gains. Is R&D failing to deliver? Bloom et al. propose to measure the effectiveness of R&D by looking at how much economic growth we get per unit of R&D on average. They find that this average has declined substantially, suggesting that R&D has become less effective, that “ideas may be getting harder to find.”

This trend seems ominous, but it is not clear that the average ratio of productivity growth to R&D is an appropriate measure of R&D effectiveness. Because many factors in addition to R&D affect productivity, it is better to measure the marginal effect of R&D on productivity. Fortunately, a large empirical literature has developed methods for estimating how much the next additional unit of R&D creates economic growth.

In a recent paper, we apply these well-established methods to micro-data from the US Census for the US manufacturing sector. Our research extends the literature by using a comprehensive sample of R&D-performing firms, including many small firms, and we study a relatively long period, from 1976 to 2018. We focus on manufacturing because productivity is reasonably well-measured in this sector. Over a variety of estimations, including estimates with instrumented R&D, we find a consistent pattern: the incremental output from an extra unit of R&D has increased substantially over the last four decades. A one dollar increase in R&D generates a much larger increase in output than it did in the past. At the margin, it appears that R&D has become substantially more effective at generating productivity-enhancing ideas, not less.

From “research” to “development”

But this deepens the puzzle: if R&D has become more effective, why has productivity growth been stagnant? Part of this paradox arises from an unrecognized implication of rising R&D effectiveness. As firms’ R&D becomes more effective, rivals’ R&D also becomes more effective. Hence, rivals invest more in R&D, they develop more innovations, and, as a result, they raise the risk that any firm’s technology will become obsolete. Using a model of R&D demand, we measure technological rivalry and the associated obsolescence rate (the speed at which innovations become obsolete). We find that both have risen sharply since 1990.

We can see evidence of rising technological rivalry and rising obsolescence in the changing nature of innovation: innovations are becoming more short-lived. For instance, product life cycles of electronic components have contracted by two thirds since 1970. This process has been hastened by rapid product proliferation; some markets are flooded with new products that use minor improvements to replace older products; from 1970 to 2012, the number of automobile models increased from 140 to 694, the number of breakfast cereals from 160 to 4945, the number of sports shoes from 5 to 3371, and so on with many other consumer products. Also, R&D has shifted away from “research” and toward relatively short lived “development”. In the 1960s, half of R&D was research; today only a quarter is, suggesting a shift to less consequential implementation of innovation rather than innovation itself. For example, in the US pharmaceutical industry, patent grants have quadrupled since 1985, but new medicines have only increased by 60 percent.

Rising obsolescence drags down productivity growth. Even if each dollar of R&D produces more ideas than before, those ideas die faster. Companies have to spend more of their R&D replacing obsolete ideas, so their stock of new and useful knowledge grows more slowly. In practice, the gains from better R&D are cancelled out by the faster rate at which innovations become outdated. Rough calculations suggest that this growing obsolescence has been cutting into productivity growth since the mid-1990s, which explains why we see more innovation but not more rapid productivity growth.

Productive firms grow more slowly

But rising obsolescence has a broader significance: it introduces an element of “short termism” in some business decisions. For instance, new technologies often involve sunk adoption costs for training, complementary investments, etc. But if the new technologies are shorter lived, firms will have less time to recoup these investments. Hence, they may forgo more marginal investments. Similarly, firms with productive technologies may be more reluctant to sink investments needed to expand. Our research shows that rising obsolescence accounts for a substantial slowdown in the growth of productive firms. The reallocation of resources to productive firms has been shown to be a major component of aggregate productivity growth – as more productive firms grow larger, the average productivity of the economy rises. Importantly, this reallocation has slowed significantly in the US over the last couple of decades; rising obsolescence accounts for most of this slowdown. Thus, rising obsolescence slows productivity growth both by slowing firm productivity growth and by slowing the expansion of more productive firms.

But we need to be clear: these detrimental effects are an unfortunate byproduct of increased innovation that ultimately increases output. Policy should not reduce incentives for innovation, but should, instead, encourage long-lived innovations rather than short-term ones. Policy can encourage long-term research relative to short-term product development; it can change patenting rules to discourage the current flood of trivial inventions and provide longer patent terms for some innovations; it can encourage governments and universities to pursue long-term basic research. R&D is not failing us, but the nature of innovation has changed; policy needs to reflect those changes.