Sonia Jaffe

We explore quality externalities on platforms: when buyers have limited information, a seller's quality affects whether her buyers return to the platform, thereby impacting other sellers' future business. We propose an intuitive measure of this externality, applicable across a range of platforms. Guest Return Propensity (GRP) is the aggregate propensity of a seller's customers to return to the platform. We validate this metric using Airbnb data: matching customers to listings with a one standard deviation higher GRP causes them to take 17\% more subsequent trips. By directing buyers to higher-GRP sellers, platforms may be able to increase overall seller surplus.

In this note we consider a scenario where agents must make a binary choice about where to locate. They have heterogeneous preferences over locations. In addition, whichever location an agent chooses, his utility also includes an agglomeration effect that is increasing in the proportion of the population that chooses that location. Each individual is uncertain about the distribution of preferences in the population, but knows his own preferences, which serve as a noisy signal of population preferences. We show that if the marginal returns to agglomeration are sufficiently decreasing in the proportion of the population co-locating (returns are more concave than log) then there will be over-agglomeration in equilibrium, relative to the social optimum. If the agglomeration function is less concave than log (for example, linear) there will be under-agglomeration in equilibrium.

We analyze the financial value of insurance when individuals have access to credit markets. Loans allow consumers to smooth shocks across time, decreasing the value of the smoothing (across states of the world) provided by insurance. We derive a simple formula for the incremental value of insurance and show how it depends on individual age, health, and income and on the features of available loans. Our central contribution is to derive formulas for aggregate welfare that can be taken to data from typical studies of health insurance. We provide both exact formulas that can be taken to data on the distribution of medical expenditures and income and an approximate formula for aggregate data on medical expenditure. Using the Medical Expenditure Panel Survey we illustrate how the incremental value of insurance is decreasing with access to loans. For consumers in the sickest decile, access to a five-year loan decreases the incremental value of insurance by $338 (6%) on average and $3,433 (36%) for the poorest consumers. We also find that our approximate formula is a reasonable proxy for the exact one in our data.

The recent rise of hybrid work poses novel challenges for synchronizing in-office work schedules. Using anonymized building access data, we quantified coattendance patterns among ~43k employees at a large global technology company. We used two-way fixed effects regression models to investigate the association between an employee’s presence in the office and that of their manager and teammates. Our analysis shows that employee in-person attendance was 29% higher when their manager was present. Moreover, a 1-SD increase in the share of teammates who were present yielded a 16% increase in the individual employee’s attendance. We also observed greater coattendance among employees who were recently hired, have a Corporate or Operations role, or work in shared office spaces. Thus, we find evidence of some voluntary alignment of work schedules. Companies could bolster such organic coordination by leveraging digital scheduling tools or providing guidance specifically aimed at increasing coattendance.

The growth in remote and hybrid work catalyzed by the COVID-19 pandemic could have significant environmental implications. We assess the greenhouse gas emissions of this transition, considering factors including information and communication technology, commuting, noncommute travel, and office and residential energy use. We find that, in the United States, switching from working onsite to working from home can reduce up to 58% of work’s carbon footprint, and the impacts of IT usage are negligible, while office energy use and noncommute travel impacts are important. Our study also suggests that achieving the environmental benefits of remote work requires proper setup of people’s lifestyle, including their vehicle choice, travel behavior, and the configuration of home and work environment.

The COVID-19 pandemic has shaken the world to its core and has provoked an overnight exodus of developers who normally worked in an office setting to working from home. The magnitude of this shift and the factors that have accompanied this new unplanned work setting go beyond what the software engineering community has previously understood to be remote work. To find out how developers and their productivity were affected, we distributed two surveys (with a combined total of 3,634 responses that answered all required questions) weeks apart to understand the presence and prevalence of the benefits, challenges, and opportunities to improve this special circumstance of remote work. From our thematic qualitative analysis and statistical quantitative analysis, we find that there is a dichotomy of developer experiences influenced by many different factors (that for some are a benefit, while for others a challenge). For example, a benefit for some was being close to family members but for others having family members share their working space and interrupting their focus, was a challenge. Our surveys led to powerful narratives from respondents and revealed the scale at which these experiences exist to provide insights as to h ow the future of (pandemic) remote work can evolve.

The coronavirus disease 2019 (COVID-19) pandemic caused a rapid shift to full-time remote work for many information workers. Viewing this shift as a natural experiment in which some workers were already working remotely before the pandemic enables us to separate the effects of firm-wide remote work from other pandemic-related confounding factors. Here, we use rich data on the emails, calendars, instant messages, video/audio calls and workweek hours of 61,182 US Microsoft employees over the first six months of 2020 to estimate the causal effects of firm-wide remote work on collaboration and communication. Our results show that firm-wide remote work caused the collaboration network of workers to become more static and siloed, with fewer bridges between disparate parts. Furthermore, there was a decrease in synchronous communication and an increase in asynchronous communication. Together, these effects may make it harder for employees to acquire and share new information across the network.

The Covid-19 pandemic dramatically changed how organizations worked. Microsoft was one of the first to ask employees to work from home (WFH). We developed an anonymous nightly diary study with 435 participants, and we learned about their experiences over the first 10 weeks of the WFH directive. We found the largest challenges were having too many meetings, feeling overworked, and physical and mental health. However, there were things to be grateful for, and many people were thankful for family, increased flexibility, their job, and their team. We also learned that the simple act of reflecting nightly during the study could be helpful: people who reported no gratitude were 22% (p-value=.000007) less likely to report being satisfied that day. Our management used the anonymized, aggregate data to create new programs (such as No Meeting Friday) to address these challenges. We then saw immediate feedback on these programs in the diaries and used that to inform future decisions.

We analyze the effects of taxation in two-sided matching markets, i.e.~markets in which all agents have heterogeneous preferences over potential partners. In matching markets, taxes can generate inefficiency on the allocative margin by changing who is matched to whom, even if the number of workers at each firm is unaffected. While the allocative inefficiency of taxation need not be monotonic in the level of the tax when transfers flow in both directions, we show that it is weakly increasing in the tax rate for markets in which workers refuse to match without a positive wage.
     We introduce a renormalization that allows for an equivalence between markets with taxation and markets without taxation but with adjusted match values. We use our equivalence to show additional properties of matching markets with taxation and to adapt existing econometric methods to such markets. We then estimate the preferences in the college-coach US football matching market and show through simulations of tax reforms that the true deadweight loss can differ dramatically from that measured without accounting for the preference heterogeneity of the matching market.
     In addition to highlighting the potential for allocative distortions from taxation, our model provides a continuous link between canonical models of matching with and without transfers.

In the past three years, there have been not one but two generational shifts in how work gets done, both of which were only possible because of decades of research and development. The first shift occurred when COVID made us realize how powerful remote and hybrid work technologies had become, as well as how much science was available to guide us in how to (and how not to) use these technologies. The second arrived this year, as it became clear that, at long last, generative AI had advanced to the point where it could be valuable to huge swaths of the work people do every day.

We began the New Future of Work Report series in 2021, at the height of the shift to remote work. The goal of that report was to provide a synthesis of new – and newly relevant – research to anyone interested in reimagining work for the better as a decades-old approach to work was challenged. The second New Future of Work Report, published in 2022, focused on hybrid work and what research could teach us about intentionally re-introducing co-location into people’s work practices. This year’s edition, the third in the series, continues with the same goal, but centers on research related to integrating LLMs into work.

Throughout 2023, AI and the future of work have frequently been on the metaphorical – and often literal – front page around the world. There have been many excellent articles about the ways in which work may change as LLMs are increasingly integrated into our lives. As such, in this year’s report we focus specifically on areas that we think deserve additional attention or where there is research that has been done at Microsoft that offers a unique perspective. This is a report that should be read as a complement to the existing literature, rather than as a synthesis of all of it.

This is a rare time, one in which research will play a particularly important role in defining what the future of work looks like. At this special moment, scientists can’t just be passive observers of what is happening. Rather, we have the responsibility to shape work for the better. We hope this report can help our colleagues around world make progress towards this goal.

This report presents the initial findings of Microsoft’s research initiative on “AI and Productivity”, which seeks to measure and accelerate the productivity gains created by LLM-powered productivity tools like Microsoft’s Copilot. The many studies summarized in this report, the initiative’s first, focus on common enterprise information worker tasks for which LLMs are most likely to provide significant value. Results from the studies support the hypothesis that the first versions of Copilot tools substantially increase productivity on these tasks. This productivity boost usually appeared in the studies as a meaningful increase in speed of execution without a significant decrease in quality. Furthermore, we observed that the willingness-to-pay for LLM-based tools is higher for people who have used the tools than those who have not, suggesting that the tools provide value above initial expectations. The report also highlights future directions for the AI and Productivity initiative, including an emphasis on approaches that capture a wider range of tasks and roles.

Due to the “Great Remote Work Experiment” that began in March 2020 when workplaces around the world rapidly shut down, work is changing faster than it has in a generation. As many people now return to the workplace and begin to experiment with hybrid work, a range of different outcomes is possible. Thankfully, researchers at Microsoft and from around the world have been investigating evolving hybrid work practices and developing technologies that will address the biggest new challenges while taking advantage of the biggest new opportunities.

This Microsoft New Future of Work Report 2022 summarizes important recent research developments related to hybrid work. It highlights themes that have emerged in the findings of the past year and brings to the fore older research that has become newly relevant. Our hope is that the report will facilitate knowledge sharing across the research community and among those who track research related to work and productivity. This research area is unfolding as rapidly as work is changing, and the purpose of this report is to help the community build on what has been learned this past year.

Never before has there been such an opportunity to actively shape the future of work. With research and careful study, we can create a new future of work that is meaningful, productive, and equitable.

The coronavirus pandemic not only caused a public health crisis, it also caused technological, social, and cultural disruption. This past year, people across the globe experienced a rapid shift to remote work that upended their existing practices and will have long-term implications for how work gets done in the future. Looking forward, we expect that some of those who used to work from offices will continue to work remotely, while others will adopt hybrid models that will involve a combination of working from the office and working remotely. The current moment presents a unique opportunity to understand the nature of work itself, to improve remote support for a range of work practices, and to use what we have learned through remote work to improve in-office and hybrid practices.

As a company whose mission is to empower every person and every organization on the planet to achieve more, it is vital that Microsoft understands the massive transition currently underway so that we can help our customers come through this challenging time stronger and more resilient. We are all right now participants in a giant, natural, uncontrolled remote work experiment from which Microsoft must learn. Just as research has been fundamental in developing ways to prevent and treat COVID-19, it is also fundamental to understanding and supporting evolving the sociotechnical work practices.

At the start of the pandemic, researchers from across Microsoft formed an ongoing cross-company initiative to coordinate efforts with the goal of understanding the impact of remote work and identifying opportunities to support new working practices. The initiative consists of over 50 research projects conducted by teams that span a range of disciplines (including engineering, research, marketing, human resources, and facilities) and divisions (including Microsoft Research, Office, Windows, Azure, Xbox, GitHub, and LinkedIn). The projects employ many different methodologies, ranging from small-scale, formative interviews with customers to large-scale modeling exercises and even EEG measurements of electrical impulses in the brain.

This report provides a synthesis of the findings from these many research projects. We believe it represents the largest compilation of research related to the pandemic's impact on work practices available to date. The findings highlight a number of acute challenges and suggest opportunities to develop new work practices that are more efficient, equitable, and energizing. Work will never again be the same. With care and effort, however, we hope to make it better.