James Bessen and Cesare Righi
Companies are making immense investments in new information technologies (IT) such as Big Data and “artificial intelligence.” Computers can now interpret X-rays to diagnose disease, select job applicants, handle routine customer support phone calls, and drive cars. There is much controversy and speculation around how these technologies will affect society. New research from TPRI provides some answers as to what is actually happening on several issues:
- Some people claim that these technologies are eliminating lots of jobs. We estimate the effect on employment at the firms investing in these technologies.
- We estimate whether the productivity growth from IT has slowed or increased in the last decade.
- We evaluate whether these technologies are associated with rising profits and a falling share of income going to employees.
This research paper uses a new methodology to distinguish the effects of firms’ major investments in developing new IT from other factors that influence revenue, profits, and employment. US firms today spend over $250 billion on developing new software, including both in-house and contracted projects. This is almost as large as net capital expenditures. Moreover, firms across all sectors are making these investments, not just tech firms.
About half of this investment takes place in major discrete projects, what we call “IT shocks.” We identify these IT shocks by using LinkedIn data to distinguish years when firms dramatically increase their employment of software developers. We can then compare the behavior of the firms making these large investments with the behavior of similar firms (for the wonks, we use difference-in-differences supplemented by a control function).
Contrary to claims that new IT is displacing workers, we find that these major IT investments are actually responsible for a 7% increase in the employment of non-IT workers on average. While automation allows machines to execute some tasks formerly performed by humans, it also appears to substantially increase the demand for the firm’s products, which increases the number of workers required. Also, we find that firms that use Big Data or machine learning exhibit even faster job growth on average. However, the employment effect is uneven: we see a decline in non-durable manufacturing employment contrasting with strong job growth in services, finance, and trade. This finding suggests that while IT and AI are not creating mass unemployment, they are destroying some jobs while creating others. The result is that workers need to be able to transition to new industries and occupations and acquire new skills. This is a major policy challenge.
Second, we find that these major IT investments improve firm productivity by about 5% on average. We also look at how those productivity gains changed over time. Robert Gordon contends that IT innovation has been exhausted, bringing smaller productivity gains in the 2000s than in the 1990s. We find that that is not so. If anything, productivity gains from more recent IT shocks are a bit larger. Nor has the frequency of these shocks declined. We do find, however, that employment growth following IT shocks has been slower after 2000. Normally, more productive firms grow larger and less productive firms shrink or exit the market. This reallocation translates firm-level productivity gains into gains in aggregate productivity. But the slower employment growth since 2000 may mean that the contribution of firm productivity growth to aggregate productivity growth is smaller. Further research is needed to understand why, but the fault seems to lie not with the technology but with the economy.
Third, we find that IT shocks are followed by a decline in the share of firm revenues that go to wages and salaries. While worker pay does not decline following the shocks, the decrease in labor’s share has raised concerns about rising inequality. Most of the decline in labor’s share appears to be from a rise in firm profits; that is, a bigger share of the benefits of the technology is going to the companies and their owners while a smaller share goes to the workers. This also raises concerns about rising market power. TPRI is conducting additional research to understand how IT shocks affect rival firms and the relative size of large firms within industries.
The study reveals many other facets of the diverse ways major IT expenditures impact firms, workers, and the economy. In this analysis, IT is much more than just another input into production. Rather it appears to be central to several critical economic issues.