QED Research Spotlight article by JDI Research Associate Brock Mutic highlighting recent research by QED Professor Sitian Liu

The COVID-19 pandemic upended our lives and the economy in unprecedented ways. Institutions and ways of organizing society which had previously been ubiquitous, had to be rethought and reworked on a dime, because of a deadly virus. One such institution was the office: although once a quintessential symbol of modern economic life, and an unquestioned premise of the modern labour market, the pandemic dissolved the office’s status as a synecdoche of work. In particular, it caused workers in unprecedented volumes and variations of jobs to transition to remote work, and to work-from-home models. The office was dead in the era of COVID-19, and Zoom had killed it.  

Although remote workers gained increased work flexibility and safety from the virus, they lost their physical connectedness to each other. Coworkers were no longer separated by floors or cubicles, nor firms by blocks or kilometers. Economic relations, while previously physical and direct, became abstract and mediated; instead of relating directly to other people, remote workers separated by an unquantifiable digital distance could only directly relate to the shadows of the people on their screens. 

To economists, the loss of physical proximity experienced by those who adopted work-from-home was an interesting detail, because close physical proximity between firms and workers located in dense urban environments has been found to boost wages and productivity levels in comparison to rural areas; in particular, density increases economic interactions, facilitates the transition of knowledge, the learning of new skills, and reduces the costs of professional networking, and fostering business relationships. Such wage and productivity boosts of density are referred to by economists as ‘agglomeration economies’. Given that urban firms who adopted work-from-home during the pandemic lost their quintessentially urban close proximity to other firms, economists thus wondered whether their agglomeration economies would be impacted; they wondered whether adopting work-from-home—at the cost of physical density—would negatively impact urban firms’ productivity and wages, which, before the pandemic, existed at the levels they did because of their nearness to others. 

To investigate whether the rise of work-from-home impacted the agglomeration economies of cities, Queen’s Economics Department Assistant Professor Dr. Sitian Liu stepped into the role of economic detective. As a labour and urban economist who studies the spatial distribution of economic activity, and the causes of wage differences across local labour markets, the question resonated with her research interests: as the pandemic was an exogenous shock that prompted a shift to work-from-home for firms in various occupations, she recognized it as a potentially fruitful natural experiment to study the effect of work-from-home on the agglomeration economies of cities.  

To solve the case, Dr. Liu, along with her collaborator Dr. Yichen Su, built a theoretical model of work-from-home (summary and Working Paper). In doing so, they were able to predict the likely impacts of working from home on the urban agglomeration effect and on the urban wage premium, as well as on aggregate economic output, before taking such predictions to the empirical data. Their model consists of an economy with an urban and a rural area, in which workers can choose to work onsite in either area or remotely work from rural areas for high-productivity urban firms. It shows that as the pandemic, as an outside—or ‘exogenous’—shock, decreased the costs associated with remote work, the number of onsite workers in urban firms decreased, because such urban workers could access urban wages without incurring urban housing costs, by physically residing in the rural area. The reduction in the number of onsite workers present in urban areas, the model demonstrated, could also possibly weaken the strength of urban agglomeration economies. 

According to the model, if urban agglomeration effects were not significantly weakened by increased work-from-home adoption and fewer workers being physically present in urban areas, then the labour supply facing urban firms would increase; that is, increasing numbers of rural workers would seek to work remotely for urban firms. This labour shift would lower urban wages, and thus the gap between the wages offered by urban and rural firms—or the urban wage premium—but increase aggregate economic output, as it would shift workers away from relatively less productive rural firms and towards more productive urban firms. 

In contrast, if reduced physical interactions as a result of increased work-from-home adoption did harm urban productivity—if urban agglomeration economies were weakened—then the model predicts that the appeal of working for urban companies would have decreased, causing some urban workers to switch over to rural firms, thereby reducing urban employment. Decreased urban productivity would again reduce the urban wage premium, but would also induce a decline in aggregate productivity and output. 

In sum, their model predicts that the increased adoption of remote work by firms in many occupations would lower the urban wage premium in these occupations, regardless of whether the prevalence of the remote work adoption weakened agglomeration effects. However, if the lowered urban wage premium was indeed driven by weakened agglomeration economies, the employment of these occupations should have shifted from large cities or urban areas toward small cities or rural areas. With their theoretical predictions, Dr. Liu and her co-author turned to the data. They sought to determine whether their predictions overlapped with the empirical impacts of remote work on urban economies, to determine whether their model captured the effects at play.  

Dr. Liu and Dr. Su’s first step was to determine how wages have changed since the start of the pandemic, among occupations that did and did not adopt work-from-home in high volumes. To do so, they drew on evidence from job listings from 2018-2023. Specifically, they collected a sample from Burning Glass Technologies, a US data firm, which spanned 40,000 job-listing sites in the US, covering approximately 70% of postings in the American economy. Ultimately, by weaving together their sample with data from the American Community Survey and the Occupational Information Network data on occupational characteristics, they found, firstly, that the percentage of companies engaging in remote work in the US skyrocketed after the start of the pandemic, and stayed high through 2023, though adoption rates varied by occupation. Secondly, they found, as their model predicted, that a statistically significant decline in the gap in pay offered by urban and rural firms had occurred for jobs that had a high degree of work-from-home adoption—the urban wage premium had declined for such jobs. The team then completed several robustness checks, and confirmed, for example, that the effect was not driven by high-wage firms merely leaving cities during the pandemic, or firms paying more to onsite workers than comparable remote workers. Having ascertained the validity of the results, Dr. Liu and her co-author concluded they provided evidence that firms’ adoption of remote work has caused a decline in the urban wage premium since the start of the COVID-19 pandemic. 

Turning to the question of the impacts on employment, Dr. Liu and Dr. Su then turned to data from the US Quarterly Census of Employment and Wages and found that for occupations with significant work-from-home adoption, there had indeed been large shifts in employment from large cities to small cities during and after the pandemic. As their model predicted precisely such a result, had urban agglomeration economies weakened, their results provided evidence of weakened agglomeration economies as a result of remote work in the pandemic era. 

Finally, to corroborate their conclusions with evidence from outside their model, Dr. Liu and Dr. Su turned to examine the job skills driving the decline in the urban wage premium. Since their original data from Burning Glass contained detailed information on the skill requirements of each job listing, the team was able to use a statistical technique known as a ‘Gelbach decomposition’ to break down the urban wage premium into the compensation offered for the various skills which constitute particular jobs; doing so allowed them to identify the skills which were driving the decline in the gap between urban and rural compensation. Ultimately, they found that the effect varied for jobs with differing levels of education: for jobs that required college degrees, the urban wage premium declined because business management skills, customer support skills, finance skills, and relationship-building skills all offered lower compensation as a result of remote work—though the effect was primarily driven by relationship skills. For jobs with no degree requirements, the effect was driven primarily by falling returns to IT skills—though also by declining returns to marketing skills and management skills. It was also found that relationship-building skills were required less intensively by high-skilled jobs located in large cities following the pandemic, and IT skills were required less intensively by low-skilled jobs in large cities. 

Since declining returns to IT skills drove the decline in the urban wage premium for jobs without degree requirements, and given IT skills can be easily transferred to remote contexts, Dr. Liu noted that the urban wage premium decline for such jobs was thus consistent with an increase in the labour supply of remote workers with IT skills reducing the returns to such skills. In contrast, since the declining wage premium for jobs with degree requirements was driven by falling returns to relationship skills, which cannot be easily transferred to remote contexts, Dr. Liu argued the effect was unlikely to have resulted through an increase in the remote labour supply. Rather, relationship skills were likely to have become less useful for production in the context of remote work in large cities. As relationship-building skills tend to be more conducive to in-person interactions, Dr. Liu thus argued that, in agreement with their other results, urban agglomeration economies—or the boosts to productivity that are induced by in-person interactions—are likely to have decreased as a result of remote work during the COVID-19 pandemic. Together, these findings indicated that remote work mostly harmed productivity in the high-skill labour market. 

Overall, Dr. Liu’s research produced insightful results and highlighted that an unintended consequence of the transition to work-from-home during the pandemic was reduced urban productivity due to lower in-person interactions in the workplace. However, full work-from-home arrangements are unlikely to last in the long run, as many firms are adopting hybrid work models. The team thus recognizes the need for future research to understand how hybrid work might mitigate the negative effects of fully remote work; a notable finding of Dr. Liu’s work is that there could be positive effects on aggregate productivity if the negative impacts of remote work on agglomeration economies could be considered when designing future hybrid work arrangements.