General Purpose Technology and Wage Inequality
GAICIL Insights
The new technology that rewards veterans sounds like a contradiction.1 This paper explains why it is not. General Purpose Technology and Wage Inequality, by Philippe Aghion, Peter Howitt and Giovanni L. Violante, tackles a neglected aspect of the link between new information technologies and rising US wage inequality. Earlier work emphasized the accelerated pace or the skill bias of capital-embodied technical progress. The authors focus on generality.
Generality gets two precise meanings. On the human capital side, a more general technology allows greater transferability of skills across sectors: using computers for word processing, databases or programming is useful in a wide range of sectors and jobs. On the physical capital side, a more general blueprint increases compatibility between old and new vintages of capital, so equipment embodying an old technology can be retooled for the new one. Skills travel; machines get a second life.
The machinery is a simple overlapping-generations model with two-period lived agents and two production sectors. Workers accumulate experience through learning by doing. Only adaptable workers can be productive with the new technology, and adaptability comes from the schooling investment of young workers, a link between education and technology adoption that goes back to Nelson and Phelps.
Result one: with adaptability fixed, more generality on either measure raises long-run wage inequality. Skill transferability raises the premium of adaptable workers through skill deepening; vintage compatibility does the same through capital deepening. In the transition, inequality may temporarily overshoot its new steady state. The spike is worse than the plateau.
Result two: with adaptability endogenous, rising generality raises the demand for education and the supply of adaptable workers, pushing back against the first mechanism. The authors characterize when the initial forces dominate, and a simple calibration suggests they do for the US economy. Schooling softens the blow without erasing it.
Result three solves a puzzle. In the United States, the ratio between the average wage of workers with 25 years of experience and those with 5 years rose by 20 percent since the early seventies. Pure accelerating-technology stories predict the experience premium should fall. A more general technology, by making skills transferable to new vintages, can raise the relative wage of older workers. Experience that travels keeps its value.
Bring the result forward to the present. Artificial intelligence is the obvious test case, widely regarded as a general-purpose technology used across sectors and occupations, much as computers were. The model implies that AI's generality could raise the premium for adaptable workers who carry their skills across applications, widening long-run inequality.
The endogenous adaptability result is the constructive answer. If generality raises the return to adaptability, it also raises the incentive to invest in education that builds it. Expanding access to such education can increase the supply of adaptable workers and moderate inequality while the economy captures AI's benefits. The policy target is adaptability itself.
Vintage compatibility matters too. AI software can often run on existing hardware and integrate with legacy systems, increasing compatibility between old and new capital. That may speed diffusion and raise the productivity of workers using the new tools. Old machines can learn new tricks.
The experience premium carries hope. When accumulated knowledge transfers to new tools, a general technology raises its value. Experienced workers who adapt to AI may find their skills worth more, a pattern that challenges the assumption that new technology always favors the young. The twenty-five-year veteran with a new tool may outearn the newcomer with the same tool.
Pulled together, a technology's general-purpose nature shapes who gains from it. Greater generality can raise inequality by rewarding adaptability, and endogenous education can push back. As AI spreads, investment in adaptability will help more workers share in the gains.
One question to put to workforce strategists: the obsolete-veteran story may have AI backwards. The paper's logic says the scarce asset is adaptability, and an experienced worker who has it holds both the knowledge and the passport to carry it into the new vintage.
Introduction
The recent changes in the US wage structure are often linked to the new wave of capital-embodied information technologies. The existing literature has emphasized either the accelerated pace or the skill-bias of embodied technical progress as the driving force behind the rise in wage inequality. A key, neglected, aspect is the “general purpose” nature of the new information technologies. This paper formalizes the idea of generality of technology in two ways, one related to human capital (skill transferability) and one to physical capital (vintage compatibility) and studies the impact of an increase in these two dimensions of technological generality on equilibrium wage inequality.
Takeaway
We have developed a simple two-sector OLG model with two-period lived agents and two-period technological vintages to analyze the e ects of an increase in the generality of technologies on wage inequality. We have modelled generality, both in relation to human capital and the transferability of skills across sectors, and in relation to physical capital and its ability to be upgraded towards other sectors using more advanced technologies. We have focused our attention on two measures of wage inequality, namely the ratio between high and the low wages within a generation of ex-ante equal workers, which we took as a measure of within-group inequality, and the ratio between the average wages of old and young workers, the experience premium. Our main ndings can be summarized as follows: First, whether it relates to human or to physical capital, an increase in the generality of technology increases long-run within-group inequality, and during the transition towards a more general technology, within-group inequality overshoots its new steady-state level. Second, the steady-state experience premium reacts positively to an increase in the generality of technologies with regard to human capital, whereas it reacts negatively to an increase in the generality of technologies with regard to physical capital; moreover, during the transition period following an increase in the former, the experience premium increases monotonically over time towards its long-run steady state.1
Third, the above ndings appear to be robust to endogenizing old workers' adaptability to new technologies provided the marginal cost of adaptability increases suÆciently fast. Fourth, a simple calibration exercise on US data suggests that the dimension of generality that increased most is that related to human capital. To obtain analytical results we had to keep our model stylized, and as such it should be seen as a benchmark model that can be extended in several interesting directions. A rst extension would explore the dynamics of wage inequality when agents live for more than two-periods. Our analysis on a related model with in nitely-lived agents (Aghion, Howitt, and Violante, 2000) shows that our approach based upon random adaptability to new technologies and its e ects on the distribution of labour market histories can account for an important observation about wage inequality, which the two-period OLG model in this paper could not capture, namely that a sizeable part of the observed increase in within-group inequality relates to the temporary component of earnings (Gottschalk and MoÆtt, 1994, Gittleman and Joyce, 1996, Blundell and Preston, 1999). A second extension, which we begin investigating in the Appendix, is to model another dimension of between-group inequality, namely the educational premium. Based on the assumption that more educated workers are also more adaptable to new technologies when old, one can show that the comparative statics across steady-states of the educational premium mirror those of within-group inequality, and in particular also the educational premium increases with the generality of technologies.
Finally, the past thirty years have witnessed economy-wide evolutions in the organization of rms, in particular a generalized move towards atter and more decentralized hierarchical forms. Recent empirical work (e.g. Caroli and Van Reenen, 2001) has pointed to the skill-biased nature of organizational change. But equally important is the generality of this change, which has a ected a broad range of sectors in manufacturing and services, and involved considerable wage premia for the most adaptable managers and labourers. This suggests that the model in this paper could also be used to analyze the e ects of this type of organizational change on the wage structure.
Notes
- Philippe Aghion, Peter Howitt & Giovanni L. Violante, General Purpose Technology and Wage Inequality, 7 J. Econ. Growth 315 (2002), https://doi.org/10.1023/a:1020875717066. Takeaway source: Concluding Remarks, reproduced from a free author or repository copy. ↩
Cite as: Philippe Aghion, Peter Howitt & Giovanni L. Violante, General Purpose Technology and Wage Inequality, 7 J. Econ. Growth 315 (2002), https://doi.org/10.1023/a:1020875717066
Header image: Hyderabad Hitech City - residential street with gate that only lets pedestrians go through.jpg by Syced, CC0, via Wikimedia Commons.
