Are We Yet Sick of New Technologies? The Unequal Health Effects of Digitalization

Authors – Melanie Arntz , Sebastian Findeisen , Stephan Maurer, Oliver Schlenker.

Abstract :

While the labor market effects of digitalization have been studied extensively, very little is known about its impact on health. Yet, theories from organizational science and psychology suggest that new technologies at the workplace may induce “technostress”. Conversely, digital technologies and automation hold the potential to relief workers from physically strenuous tasks.
In this paper, we use a novel employer-employee data set that combines survey and administrative information from social security records and allows for examining the link between the changing use of frontier technologies at the workplace between 2011 and 2019 to individual health outcomes.
In the first step of our analysis, we document that new technologies that diffuse into workplaces in the 2010s are mainly the Internet of Things, the Internet of Services, but also big data and artificial intelligence tools, virtual augmented reality, and blockchain technologies.
We then use a measure of individual exposure to an increased use of these frontier technologies and allow related effects to depend on someone’ initial task structure. For this, we run a principal component analysis for over a dozen tasks reported for the initial workplace to reduce the dimensionality of the task space. It turns out that most initial variation in task profiles is captured by one single component that we name Office Score (OSC) that tends to increase with the frequency of office-related, more complex, and non-manual tasks. If workers at the lower end of this distribution are less used to complex tasks, the increased use of frontier technologies may result in more technostress and worse health outcomes.
We estimate these health effects using a first-difference model that relates individual-level change in technology use to individuals’ health while accounting for a rich set of controls, including individual characteristics as well as occupation, industry, or firm fixed effects.
For the average workers, we find economically and statistically insignificant health effects that mask substantial effect heterogeneity: For workers starting in predominantly manual and service-oriented jobs, i.e.~in the lowest tercile of the office-score distribution, a one standard deviation increase in the treatment variable is associated with a robust decline in reported subjective health of 0.115 standard deviations and an additional 7.5 days of sickleave. These effects dissipate for those workers who participated in employer offered training programs, suggesting that active policies at the firm level might be an effective way to assist workers in adjusting in response to technology shocks.