Category: Data Source

Science seems like a public good; in theory, ideas are non-rivalrous and non-excludable. But the closer we look at how ideas actually spread and are used in the world, the less they seem like public goods. As I am fond of pointing out, Thomas Keller wrote a literal recipe book for the dishes he served at his world famous French Laundry restaurant and yet, the French Laundry did not go out of business. Ideas are in heads and if you don’t move the heads, often the ideas don’t move either.

In a new NBER working paper, The Effect of Public Science on Corporate R&D by Arora, Belenzon, Cioaca, Sheer & Zhang, (Tyler mentioned it briefly earlier) the authors make a similar point:

…the history of technical progress teaches us that abstract ideas are also difficult to use. Ideas have to be tailored for specific uses, and frequently, have to be embodied in people and artifacts before they can be absorbed by firms. However, such embodiment also makes ideas less potent sources of increasing returns, turning non-rival ideas into rival inputs, whose use by rivals is easier to restrict. Our findings confirm that firms, especially those not on the technological frontier, appear to lack the absorptive capacity to use externally supplied ideas unless they are embodied in human capital or inventions. The limit on growth is not the creation of useful ideas but rather the rate at which those ideas can be embodied in human capital and inventions, and then allocated to firms to convert them into innovations.

The question of whether science is a public good is not merely technical but has significant implications. If science is a public good, markets will likely underproduce it, making government subsidies to universities crucial for stimulating R&D and economic growth. Conversely, if ideas are embodied and thus closely tied to their application, government funding for university research might not only fail to enhance economic growth but could also hinder it. This occurs as subsidies draw scientists away from firms, where their knowledge directly contributes to product development, towards universities, where their insights risk becoming lost in the ivory tower. (Teaching scientists who then go on to careers in the private sector is much more likely to be complementary to productivity growth than funding research which pulls scientists away from the private sector.)

In a commentary on Arora et al., the Economist notes that growth in universities and government science has coincided with a slowdown in productivity.

Universities have boomed in recent decades. Higher-education institutions across the world now employ on the order of 15m researchers, up from 4m in 1980. These workers produce five times the number of papers each year. Governments have ramped up spending on the sector. The justification for this rapid expansion has, in part, followed sound economic principles. Universities are supposed to produce intellectual and scientific breakthroughs that can be employed by businesses, the government and regular folk. Such ideas are placed in the public domain, available to all. In theory, therefore, universities should be an excellent source of productivity growth.

In practice, however, the great expansion of higher education has coincided with a productivity slowdown.

Arora et al. present detailed empirical evidence causally linking the productivity slowdown to the expansion of government science. Government science has yielded smaller-than-expected productivity improvements due to significant trade-offs. Subsidies have moved heads out of firms and into universities and for many firms this shift of talent has not only reduced the firms’ capacity to generate ideas (crowding out) but has also impaired their ability to adopt academic innovations. As the authors write:

…productivity growth may have slowed down because the potential users—private corporations—lack the absorptive capacity to understand and use those ideas.

The great Terence Kealey made many of these points much earlier in his important book, The Economic Laws of Scientific Research (here is an online precis). Kealey, however, was challenging a beautiful theory, supported by the great and good of the economics profession, by pointing to an ugly practice. Arora et al. show that the beauty of the theory may have misguided us and that “the vast fiscal resources devoted to public science…probably make businesses across the rich world less innovative” (quoting the Economist).

Greg Clark may well be the most important social scientist of the 21st century. His use of historical data informed by evolutionary theory and genetics is a unique contribution to social science with important and challenging results.

Clark’s latest paper (with Neil Cummins) makes a simple but striking point. If the primary systematic determinant of social outcomes is genetic then we expect the father and the mother to contribute equally (each giving half their genes). If, on the other hand, the primary determinant is social then we expect widely different mother-father contributions in different societies and at different times and for different characteristics. Fathers ought to matter more in patriarchies, for example, and mothers more in matriarchies and gender-egalitarian societies. Similarly, if social factors are determinative, we would surely see a rising contribution of mothers to child outcomes as the social power of women rises (you can’t use your mother’s contacts in the legal profession to get a job, for example, if your mother was never a lawyer.) Similarly, if social factors are determinative we would expect mothers to be more important perhaps for characteristics determined early and fathers for characteristics determined late.

As Clark and Cummins write:

Social institutions and conventions would suggest that social status will often be more strongly transmitted between generations on either the patriline or the matriline. The factors favoring stronger transmission on the matriline are the much greater involvement in all societies of mothers in the care and education of children. The greater time investment of mothers in childcare is found in all societies, even those such as in contemporary Nordic countries where gender equality is the most advanced. Thus we would on the human capital interpretation of social outcomes expect a greater maternal than paternal connection in the modern world. However, a countervailing force in earlier times was the greater access of fathers to resources, and professional contacts. Also since in earlier years only fathers had occupations and educational qualifications, the father could be much more of a model for the outcomes of sons. It is thus uncertain whether the paternal or maternal line would better predict social outcomes in any earlier society. But we would expect the paternal effect to be greater in high status groups, and the maternal effect greater in average or lower class families.

What we find with the FOE data, however, is that in 27 out of 31 child outcomes (other than wealth) examined across marriages in the years 1754-1995, the patriline and matriline had a predictive ability for child outcomes that was not statistically distinguishable at the 5% level. In the four cases where the coefficients differed significantly, in three the maternal effect was greater, and in one the paternal effect. Thus for most social outcomes – literacy, age at beginning work, age at leaving schooling, higher education, and occupational status – mother and fathers appear always to contribute roughly equally. The one clear exception is wealth, where always patriline wealth is a much stronger predictor of child wealth than is matriline wealth.

…The results suggest, however, that the mechanism of transmission is largely independent of parental time interacting with children. The results reported above are thus consistent with the finding of Clark (2023) that the pattern of inheritance of most social outcomes in England 1600-2022 was consistent with direct additive genetic transmission. Such transmission would imply a symmetry of mother and father predictive effects.