By Massimo Del Gatto
Labour productivity in OECD countries has grown by 3.7% since 1990 compared with 28.4% in emerging economies (OECD, 2015). Against the 7.5% growth in the United States, the rate peaked at 63.8% in China and 38.4% in India and, notably, turned into negative in some industrialized countries (-1.3% in Italy, from an annual growth of about 5% in the fifties).
These trends have sparked widespread interest in quantifying labour productivity and analysing its determinants. For instance, recent OECD work reveals that part of the labour productivity gap across countries disappears once the differences in how countries measure labour input are taken into account. In the case of the United Kingdom, the gap with the United States shrinks from 24% to 16% (see the work by Ward et al., 2018).
Technological change is usually regarded as one of the main productivity drivers and measuring the gains associated with the digital transformation is recognised as being one of today’s trickiest challenges. However, researchers have little to say about how much of the cross-firm labour productivity differentials is driven by technological differences and how much, instead, has to be traced back to the firms’ ability (i.e., workers’ skills, management practices, organizational quality) to use the actual technology.
Since labour productivity naturally grows when capital accumulates more than labour (i.e., capital deepening), conventional statistics increasingly rely on Total Factor Productivity (TFP) indicators to quantify the output differences not related to using a different amount of inputs. Such measures are very useful for disentangling the contribution of labour and capital accumulation in explaining productivity growth, but they say nothing about the ability to exploit a given technology.
In work presented at the 2018 ESCoE Conference on Economic Measurement, held at the Bank of England, and published in an ESCoE Discussion Paper today, Michele Battisti, Filippo Belloc and I suggest a novel approach to TFP estimation, which is suited to understanding to what extent the labour productivity differentials across firms are driven by two key dimensions: adoption of different technologies and ability to exploit the given technology.
Applied to a sample of about 35000 firms distributed across 22 sectors and 76 countries, this methodology shows that both dimensions play an important role: the labour productivity gap between the firms at the technological frontier and those using less productive technologies amounts to around 34% on average; the labour productivity of the top 5% firms in terms of ability to use their actual technology is on average around 31% higher than that of the other firms.
The results of the analysis confirm what we instinctively know from every day experience and observation: some sectors are much better than others at utilising new technologies and within each sector there is also huge variation among firms in how they deal with technological innovation.
What is the message then for policy-makers?
In a 1987 New York Times column, Robert Solow noticed that “you can see the computer age everywhere but in the productivity statistics”, remarking on the circumstance that America, as most industrial economies, apparently failed to capitalize on the Programmable Automation Revolution and the rapid increase in the use of IT, which indeed coincided with the general productivity slowdown that occurred in the early 1970s: this was known as the IT-productivity paradox.
After two decades, advanced economies are now facing a similar paradox (as highlighted by Brynjolfsson et al. 2017): we now live in a digital age but productivity slows down again in the face of the expected productivity gains associated with the implementation of artificial intelligence solutions (i.e., machine learning, predictive maintenance, etc.). While mismeasurement of outputs/inputs and lags due to learning and adjustment were among the most appealing suggested explanations for the first IT-productivity paradox, my analysis highlights that, now more than then, productivity gains can differ, even substantially, depending on firms’ abilities.
In terms of policy (particularly, innovation policy) prescription, this entails that there is no one-size-fits-all solution for enhancing productivity. Or at least, facilitating technology upgrade is not the only (certainly not the best) way of boosting productivity when the costs associated with technological upgrading are likely to be large. In those cases, the adoption of new technologies has to proceed hand in hand, and eventually follow, the implementation of effective policies aimed at improving workers’ skills, management practices and organizational quality, that is firms’ capacity to take fully advantage of technological progress. Understanding these dimensions is a new challenge for economists and practitioners.
ESCoE blogs are published to further debate. Any views expressed are solely those of the author(s) and so cannot be taken to represent those of the ESCoE, its partner institutions or the Office for National Statistics.
