Innovation at the Office
Innovation at the Office
Very close physical proximity helps people discover each other, if they wouldn't otherwise
Before Today’s Post, an Announcement: The Institute for Progress is hosting a free 6-week online PhD course titled “The economics of ideas, science and innovation.”
I’m teaching one of the sessions, and Pierre Azoulay, Ina Ganguli, Benjamin Jones, and Heidi Williams are teaching the rest. An all-star lineup! The course is aimed at economics PhD students who want to learn more about the economics of innovation, but we’re also open to applications from PhD students in related fields or recent graduates.
The course starts November 1, but the deadline to apply is September 6. Learn more here!
Now for your regularly scheduled content…
Like the rest of New Things Under the Sun, this article will be updated as the state of the academic literature evolves; you can read the latest version here.
You can listen to this post above, or via most podcast apps: Apple, Spotify, Google, Amazon, Stitcher.
For decades, the office was the default way to organize workers, but that default is being re-examined. Many workers (including me) prefer working remotely, and seem to be at least as productive working remotely as they are in the office. Remote capable organizations can hire from a bigger pool of workers than is available locally. All in all, remote work seems to have been underrated, relative to just a few years ago.
But there are tradeoffs. I’ve written before that physical proximity seems to be important for building new relationships, even though those relationships seem to remain productive as people move away from each other. This post narrows the focus down to the office. Does bringing people together in the office actually facilitate meeting new people? (spoiler: yes) But I’ll try and get more specific about how, when, and why this happens too.
One aside: this is a rich literature that goes back decades. I’m going to focus on relatively recent research that looks at scientists and startups and uses experimental and quasi-experimental approaches. But a lot of this recent work turns out to echo what earlier studies found using more observational approaches. Allen and Henn (2007) provides one overview of some of the older literature.
Academic Collaboration Among Neighbors
Let’s start with buildings. Are people more likely to work together on a project if they also work in the same building?
Miranda and Claudel (2021) look at what happens to collaboration between MIT-affiliated professors and staff when they start working in the same building (or get separated), due to a series of renovation and new building projects over 2005-2015. Every year they look at each pair of 1,417 MIT authors to see if the authors’ offices are in the same building, and if they were coauthors on a paper. They want to estimate the impact of being in a building together, which presents a bit of a challenge. We might expect people to seek out offices in the same building as their expected collaborators, but they would have ended up working together whether they succeeded in getting colocated offices or not. That could overstate the impact of being in the same building. So Miranda and Claudel try to estimate the impact of being in the same building, after you adjust for a particular pair of author’s underlying propensity to collaborate regardless of location.1 Essentially, pick a random pair of MIT coauthors and identify two years where they had the same number of publications in the previous year. If they were in the same building in one of these comparison years and not in the same building in the other, they tended to publish an extra 0.004 papers together in the year they were in the same building.
An extra 0.004 papers might not seem like much, but that’s because most random pairs of MIT scientists do not put out any papers together in a given year. With 1,417 MIT authors, there are over a million possible ways for them to pair off, but they only put out 38,000 papers written by multiple MIT authors collectively over the decade. That works out to about 0.004 papers per pair per year, which implies moving people into the same building about doubles the number of papers they might be expected to put out together.2
That’s about the same order of magnitude found by Catalini (2017). Catalini focuses on the Université Pierre-et-Marie-Curie and it’s 17 year quest to remove asbestos from its buildings. Asbestos removal required moving labs to new locations, typically based on what space was available rather than as a way to make inter-lab collaboration easier. Catalini also finds when labs are moved into the same building, they put out 2.5-3.3x as many joint publications as pairs of labs that are not moved together.
Going Inside the Building
That’s for two people (or groups) working in the same building. But buildings can be pretty big. What if we look within the building; do we see similar effects for people with offices that are closer or farther away from each other?
Roche, Oettl, and Catalini (2022) peers within a US co-working space that hosted 251 different startups over 2014-2017. Whereas Miranda and Claudel (2021) and Catalini (2017) needed to try and convince us that building moves were basically random due to renovation, in this case the startup residents actually were randomly allocated to different places in the co-working hub. Very convenient for the researchers!
A difficulty is startups do not typically collaborate on easily observable projects like scientific papers though. Instead, Roche, Oettl, and Catalini look for evidence that the startups trade information using data from BuiltWith that describes which web technologies startups use. For example, NewThingsUnderTheSun.com is in the BuiltWith dataset, and it shows I use CloudFlare for a bunch of stuff, and that I registered the domain name from Tucows. Suppose I moved into a coworking space with a bunch of startups that used a web technology called Mixpanel for A/B testing. Roche and coauthors can see this in their dataset. If I started using Mixpanel myself to do A/B testing for NewThingsUnderTheSun.com after moving into the coworking space, then that suggests I learned about Mixpanel from some of the other startups there.
Roche, Oettl, and Catalini measure the shortest walking distance between each pair of startups on the same floor (walking distance is the shortest path you could actually walk, respecting walls, furniture, etc) and then they look at the probability startups adopt each other’s component web technologies. As you might expect, the closer two startups workspaces are, the more likely they are to use each other’s stuff. What’s perhaps a bit surprising though is that the effect of distance is highly nonlinear. Divide the startup pairs into four groups, based on their proximity, and you find only the 25% that are closest exhibit any knowledge sharing. It looks like being in the same building only matters if you are actually really close - like, within 66 meters!
This echoes a common finding in some of the older literature I alluded to earlier. Proximity matters, but for most people the value of proximity falls off very fast. If you have to walk very far to talk with a colocated coworker, then that coworker might as well not be colocated.
Hasan and Koning (2019) get similar results in the context of a startup bootcamp in India. They randomly assign 112 aspiring entrepreneurs to 40 different teams, whose location in a large open co-working space is also randomly assigned. Bootcamp attendees spent their first week developing a project that was later evaluated by the team, and Hasan and Koning study how proximity between teams affected their interactions during this week. To measure interactions, they survey people after a week (do you know this person? Did you ask them for advice?) and also see if they sent each other more messages via email or Facebook. As with Roche and coauthors, the impact of very minor distances seems to matter a lot. The probability bootcamp attendees reported they knew, sought advice from, or frequently messaged people on other teams dropped rapidly as distance increased (focus on the black lines below, for now - we will discuss the dashed ones shortly).
It’s also worth noting that all the teams in this study were as close as the teams in the first quartile of the Roche, Oettl, and Catalini (2022) study, so even among the top 25% closest startups, it seems likely the very closest exchanged most of the information. And note, in both of these studies, the locations of teams was random - it’s not as if people were grouped by the similarity of their work. And yet, proximity seemed to matter quite a bit for information sharing anyway.
Communication or Discovery?
So far, we’ve found evidence that jamming people together in a building increases the probability that they exchange information and start joint projects, especially if their workplaces are very close within the building. This could be for at least two different reasons though.
First, being close might make it easier for people to communicate. We know this is true, in the sense that you literally don’t have to walk so far to talk face-to-face with someone who is nearby. If face-to-face conversation is a much better way to trade information than digital messaging, then we expect close coworkers to trade more information. They might also decide to start more scientific projects together, because they know it’ll be easier to complete those projects when it’s so easy to communicate. Call this the communication advantage of proximity.
Second, being close might make it easier to meet new people. You might not march across the room to introduce yourself to someone you don’t know, but you’re pretty likely to introduce yourself to strangers you are sitting next to every day. You also are more likely to overhear each other’s conversations, and be forced to make small talk while in the same general space. Let’s call this the discovery advantage of proximity, since proximity helps you discover people you didn’t know about.
Of the two, the communication advantages of proximity are seen as so obvious that to some degree it’s not even worth studying. That’s implicitly implied in Catalini (2017), Roche, Oettl, and Catalini (2022), and Hasan and Koning (2019), because in all three papers the fundamental work unit is always kept together, whether that’s a lab, startup, or bootcamp team. It’s taken as a given (or at least not examined) that you would want to make it easy for a team to communicate by keeping them together. But granting that a core team is going to be kept together, there is still the question of how you group teams. Should you sit similar teams near each other? Different teams? Does it matter?
The relative strength of the communication and discovery advantage matters here. If the communication advantage is the most important thing, then you would want to group together teams who need to work together frequently, since that makes it easier for them to work together efficiently. If the discovery advantage is the most important thing, then you would want to group teams who are unfamiliar with each other together, since proximity will help them discover each other. There’s a tension between these two, because teams who need to work together frequently would certainly discover each other in the course of working together, whether or not their workstations are nearby. On the other other hand, people who need to discover each other might end up working together but they won’t know that in advance. Communication favors grouping people together who know each other, discovery favors grouping people who don’t know each other.
Taking it as given that a core team will be colocated, all of these papers provide evidence that after that, the discovery advantage is the stronger of the two.
Catalini (2017) provides the sharpest evidence of this. He points out that if the communication advantage is paramount, we should see a decline in collaboration when teams that were previously close are separated, since separation makes it more challenging to work together. In contrast, if the discovery advantage is paramount, then there should be an asymmetry, where being placed next to new labs encourages collaboration, but being separated does not decrease collaboration, since you remain aware of people. This asymmetry is what he observes: separation has only a modest effect on subsequent collaboration, while being moved into the same building has a large impact on collaboration.
Another surprising implication of the communication advantage is that being moved together should lead to more lowerpotential projects being attempted. This is because projects with high potential benefits are more likely to be worth doing, even if communicating is harder. It is the low potential benefit projects that are not attempted when inter-lab communication is hard, and which might be attempted when proximity makes communication easier.
But we don’t see much evidence of that either. In science, it’s common to use the citations received by a paper to measure its impact,3 and so one way to check this is to see if labs that become colocated start doing more work that is relatively poorly cited. In fact, Catalini finds the opposite! Labs that were previously separated and become colocated put out more highly cited work.
Another difference between the communication and discovery advantages is that we should expect the discovery advantage to be strongest among groups that would otherwise remain unknown to each other, whereas the communication advantage is at best neutral on this front, or potentially strongest among groups that are most likely to know each other (and hence to want to work together). But in fact, we typically find the impact of proximity is strongest among groups that are likely to remain unknown to each other, in the absence of proximity.
Discovering New People and New Ideas
Each of these papers documents this in different ways.
Catalini (2017) creates a measure of the similarity of labs to each other, based on either the citations their articles make or the keywords attached to the articles they write. The more two labs cite the same articles or share the same keywords, the more similar they are. Catalini finds the increase in collaboration between labs is entirely driven by labs that are more different from each other, rather than more similar. In other words, when two labs that work on similar scientific topics are colocated, they are not more likely to collaborate than if they are separated. We can imagine these labs already knew of each other’s work, and hence closer proximity was redundant, in terms of helping them to discover each other. Meanwhile, advantages from easier communication don’t seem to have mattered enough to induce more collaboration.
Roche, Oettl, and Catalini (2022) use several different metrics to assess whether two startups were likely to be aware of each other. For example, analogously to Catalini’s classification based on the articles cited by each lab, they look at the overlap in the web technologies used by different startups. If two startups use very similar tech stacks, that might indicate they are in a similar business or draw on advice from a similar pool of people. As another indicator of higher potential to already know each other, Roche and coauthors also look at pairs of startups in the same produce market. Lastly, Roche and coauthors look at startups with a majority female team, since these groups might have tighter networks by virtue of being underrepresented among startups.
In all cases, the relationship between adopting another startups web technology and physical proximity was weakened, and usually broken, in the presence of these alternative forms of closeness. As with labs, it seems that if two startups are likely to already be aware of each other, because they work in the same industry or for other reasons, then physical proximity is redundant and doesn’t help them discover each other, nor offer much of an advantage in terms of easier communication.
Hasan and Koning (2019) provide even sharper evidence of the importance of pre-existing knowledge of peers. Unlike the other papers, they don’t have to guess if people know each other; before the bootcamp started, they asked everyone if they knew any of the other attendees! About half of the attendees already knew at least one other bootcamp attendee, before starting. This group was much less affected by proximity than those who did not know anyone. Below, I reproduce the charts from Hasan and Koning, but now look at the difference between the solid black line, which corresponds to people who entered the bootcamp alone, and the dashed line, which corresponds to people who entered the bootcamp knowing someone else.
Attendees with pre-existing ties were still more likely to say they knew people who were assigned workstations near to them, but the effect is a lot weaker than those who entered the camp alone. And they appear no more likely to ask them advice, and significantly less likely to send them digital messages.
As an additional piece of evidence, Hasan and Koning look at the correlation between the score on first week team projects and the scores of nearby team projects. If you came in not knowing anyone, the score you got was more correlated with the scores of your neighbors: if you had the good luck of getting placed next to a bunch of good teams, you ended up with a higher score, while if you had the bad luck of getting placed next to a bunch of bad teams, you ended up with a lower score. But if you came in already knowing some people, your score was basically unrelated to the scores of your nearby teams. Again, that’s evidence close proximity was a channel for meeting people and then trading information, but apparently people who already had a social network didn’t do this.
Finally, it’s worth mentioning that Miranda and Claudel (2021) find results that are also consistent with discovery being more important than communication advantages; pairs of MIT professors and staff who move to buildings that house more departments are more likely to begin a collaboration than those who are moved into buildings with fewer departments (holding fixed the number of people and size of the building). We might expect people to know most of the people in their own department, whether they are in the same building or not, but require proximity to discovery people in other departments.
All this is also broadly consistent with the emerging evidence that people are quite productive working with their existing teams when working remotely. That suggests the communication advantage of proximity has been at least partially slain by better digital collaboration technology. Some of the same literature, however, also finds that remote workers are more siloed and less apt to form connections across teams. That’s also consistent with the notion that offices help people meet who would not otherwise do so in the course of their work.
Maximizing New Contacts: Common Spaces
Suppose you wanted to go all in on the relationship creating function of the office. How would you organize your office?
Unfortunately, it seems you can’t just scale up the office to be super large to house lots of potential connections. When we peer inside of buildings, we learn that the effects of colocation are narrowly constrained; if someone is more than 100 meters away, it may not matter that they are in the same building.
So it pays to be a bit intentional about who you put close together. To maximize discovery, you should group people who are unlikely to know each via other avenues, but who might plausibly benefit from being able to share ideas. As I’ve discussed elsewhere, we have some evidence that new knowledge is most likely to be useful when it is adjacent to what you know, rather than identical or very distant. So putting together people who don’t know each other but work on somewhat related stuff but not identical stuff might be ideal. That said, sometimes, important new breakthroughs happen when connections are made between seemingly unrelated ideas.
But there are also ways to expand your proximity budget a bit. In addition to the walking distance between pairs of startups, Roche, Oettl, and Catalini (2022) also look at what happens when a common space, such as a group kitchen, open sitting space, or elevator waiting area lies along the shortest walking path between two startups. They find these common areas effectively double the distance at which startups influence each other. Whereas only the 25% closest startups exert a statistically detectable influence on each other, the 50% closest do if there is a shared common area linking them.
That’s consistent with some non-experimental evidence too. Appel-Meulenbroek et al. (2017) studies 138 R&D workers at a major Dutch document management and printing firm by asking them to keep diaries on all their workplace interactions. Among workers whose workplaces were within a 30m walk, more than 80% of informal unplanned meetings took place at one of the two employees’ workplaces. For those who were farther apart, less than 60% of these meetings took place at workplaces. Instead, the hallway or “project rooms” were much more likely to be used.
Kabo et al. (2014) also looks at the probability two scientists initiate a research project as a function of both their walking distance and their “path overlap.” The former we’ve seen in all the preceding studies. The latter measures the amount of floorspace two scientists share in their typical work routines (going to the elevator, the restroom, entering, and exiting, etc.). It might be that two people have offices that are quite far apart, but that they are still very likely to cross paths on a long walk to the elevator (or maybe the kitchen). Indeed, Kabo and coauthors finds pairs of scientists are more likely to initiate research projects when they are closer, but also when they have more path overlap.
So offices can also spark connection between people who are not very close to each other, so long as there are other aspects of the building design that funnel distant people into the same space. That might be common areas that serve a large group, but it could also be shared hallways and corridors. For example, if you forced everyone to use one big hallway that connected all parts of the office.
Maximizing New Connections: Time
Another way to force more mixing is to shift people around, so that their nearest neighbors change from time-to-time. Indeed, both Miranda and Claudel (2021) and Catalini (2017) rely on worker moves to identify the effect of colocation, so we know this works. Of course moving people around all the time could get pretty annoying, especially for people who like a good routine or find meeting new people to be draining. But occasional moves might be worth it if the communication advantages of colocation are small relative to the discovery advantages.
The move to hybrid work arrangements suggests one way to encourage this kind of frequent rotation of neighbors, so long as people’s schedules are allowed (or forced) to vary and not become synced up with the same cohort. Hybrid work might also make it less annoying to be moved frequently, since workers could enjoy routines and a break from meeting new people on their work-from-home days.
To push this line a bit farther, we could begin to ask how frequently is it necessary for workers to be in the same place to reap most of the benefits of discovery. Could a fully remote company enjoy most of the benefits of an office (from the perspective of facilitating new connections) with an annual retreat?
The literature on academic conferences strongly suggests temporary meetings help facilitate new collaboration. Joint attendees at a conference are 10-20% more likely to collaborate, compared to controls. While this is significant, it is much smaller than the estimated effects of colocation in the same building; recall Miranda and Claudel (2021) and Catalini (2017) both find colocated workers are twice as likely to collaborate, possibly more. Moreover, Catalini finds the probability of collaboration steadily rises with each year that two labs are in the same building, though Miranda and Claudel find the effect plateaus after two years. On the other hand, the academic conference literature finds much larger effects - possibly the same order of magnitude as colocation - when people meet in small groups at conferences. So it may be that short work retreats intentionally designed to foster new connections are able to match most of the benefits of passive office-based network formation. I expect we’ll see plenty of research on this in the years ahead.
As always, if you want to chat about this post or innovation in generally, let’s grab a virtual coffee. Send me an email at mattclancy at hey dot com and we’ll put something in the calendar.
New Things Under the Sun is produced in partnership with the Institute for Progress, a Washington, DC-based think tank. You can learn more about their work by visiting their website.
In technical terms, they include pair fixed effects (which captures things like common research interests or complementary skill sets, if they don’t change over time), and a lagging indicator for collaboration in the previous year (which imperfectly captures changes in things like research interests). Note their framework is essentially a two-way fixed effects difference-in-difference model, which we now know can give bad results in some settings. In this case, there is some evidence the impact of being in the same building has a one-off level effect that does not vary over time, and two-way fixed effects seems to work ok in that setting (Baker, Larcker, and Wang 2022)
Note this calculation is not part of the paper, but comes from me.
For more on the use of citations to evaluate ideas, see Do Academic Citations Measure the Impact of Ideas?