Skip to main content
SearchLoginLogin or Signup

Review 1: "Infection Risk at Work, Automatability, and Employment"

This preprint develops a model to examine the elevated risk of automation due to COVID-19 restrictions and illness sequelae. In particular, it looks at the differential risk for automation that different industries face.

Published onNov 26, 2023
Review 1: "Infection Risk at Work, Automatability, and Employment"

RR:C19 Evidence Scale rating by reviewer:

  • Potentially informative. The main claims made are not strongly justified by the methods and data, but may yield some insight. The results and conclusions of the study may resemble those from the hypothetical ideal study, but there is substantial room for doubt. Decision-makers should consider this evidence only with a thorough understanding of its weaknesses, alongside other evidence and theory. Decision-makers should not consider this actionable, unless the weaknesses are clearly understood and there is other theory and evidence to further support it.

***************************************

Review:

The paper examines the tradeoff between concerns about automation and the risks of viral transmission. The authors have developed a model to characterize this tradeoff and subsequently tested it using Austrian employment data during the Covid-19 pandemic. Employing a differences-in-differences-in-differences (dif-dif-dif) analysis, they find a negative effect on occupation-level employment stemming from the higher viral transmission risk in occupations susceptible to automation during the COVID years.

The authors utilize the viral transmission risk (VTR) developed by Chernoff and Warman (2023) and the routine task index developed by Autor et al. (2003), which serves as a proxy for occupations at risk of automation.

The paper contributes to the existing literature, which posits that the pandemic and the necessary containment measures accelerated automation—referred to as "Forced Automation" (e.g., Ding and Molina (2020), Egaña-delSol, Cruz, Micco (2022), Flisi and Santangelo (2022)).

The authors expand upon Krenz et al.'s (2021) model, introducing the capability for a negative relationship between labor productivity and the infection risk in the workplace ([atlw(itlw)]/[d itlw] < 0). In their model, firms exhibit indifference between the utilization of human labor and robots (condition [atlw(itlw)]/atpw = wtl/yryr ) . Therefore, an increase in the risk in the workplace, due to COVID for example, reduces labor productivity, hence, it induces firms to automate this task.

They test this prediction on employment (or hours worked) using the Austrian microcensus for the period 2015-2021, which includes information from about 22,500 (randomly selected) households in Austria on a quarterly basis. They sum employment across regions-sectors and occupations to study the impact of Covid-19.


Main Comments

The primary contribution of the paper lies in the evidence the authors present regarding the negative relationship between employment growth and occupations at risk of automation during the pandemic in the Austrian labor market. The key point of the paper is the amplification effect of the risk of infection on the automation process (in the econometric model, the “interaction” between ROUT and VTR indices).

 The model, the authors developed, does not provide any new insight because its main result comes directly from the assumption they made about labor productivity and risk of infection (a´<0.).

The authors use a level-panel data. The unit of observation is employment at the region-industry-occupation across quarters. The econometric models include industry, occupation, and region fixed effect. Given that the primary focus of the paper is to observe the change in the evolution of automation due to Covid-19, the authors should include:

  1. Region-Industry-Occupation fixed effects.

  2. Time fixed effect, or a trend variable.

In the econometric model, the only variable that changes over time is COVID, therefore it is difficult to see whether it captures a pre-trend condition or the effect of the COVID itself. It would be interesting to see the evolution of a set of time dummies interacted with VTR. These dummies should fall after 2019q4, mainly for occupations at risk of automation. 

For the model with all occupation, the authors should include an RTIxVTRxCovid interaction term. The models should also control for other factors discussed in the literature, like telework (Telework x Covid) and childcare (women x Covid).

In the data description section, the authors should show and discuss the correlation between RTI, VTR, Telework and Women. It is important to see how correlated these variables are to have an initial insight about how they should interact during Covid-19.

It would be interesting to see the evolution of employment during the recovery phase. The main idea of the model is that Covid accelerates the automation process. There are new studies that claim that employment in occupations at risk of automation have, by definition, a large elasticity of substitutions with robots or ICT capital, and therefore after a negative demand shock employment should fall more but recover faster after the shock (see Egaña and Micco (2022)).


Other Comments:

  1. Report the crosswalk between the International Classification of Occupations 2008 (ISCO-08) at the 3-digit level comprising 130 occupations, and Standard Occupational Classification (SOC) System –year-. And discuss why we should expect that the Routine Task Index (RTI) in the USA should be the same in Austria.

  2. In the data section, the authors should provide a description of the number of sectors, regions, and occupations employed in computing employment. It appears that there are cells or rows with very few observations, such as those exclusively with women or employees in a specific age group. If this is the case, the authors should discuss their approach to handling such instances.

Comments
4
?
wan yu:

Especially when working in the injection molding industry, the air is filled with the smell of plastic.

?
jack ferals:

Navigate cars through gridlocked streets to clear the path. Traffic Jam 3D

?
mike jack:

Avoid letting emotions dictate your bets; make rational decisions based on facts rather than personal biases or preferences. free betting tips

?
jack ferals:

Mario's iconic abilities, such as jumping, running, and stomping enemies, are key to navigating through each level and overcoming challenges.