The Brink: What made you want to study job applications on online platforms?

Fradkin: One of the most important markets in the world is the labor market. People need to find jobs. And as economists, we know that these markets have various inefficiencies. There’s a potential by designing these platforms to prevent this from happening, to make the market more efficient. So, the way I think about it is: making a more efficient jobs platform is going to hopefully result in people being matched with better jobs, but also less wasted time. 

Now, whether we can get there is a different question. In this paper [we were asking]: When you present different types of information as people are searching for a job, how did people react to that? The other interesting part of it was that Meta had this jobs platform called Jobs on Facebook, and they let us collaborate with them to run an experiment and to study it. And, as an academic, studying a real-world platform where we can design the experiment is really cool and unique.

The Brink: How does the number of applicants for a role impact the way that people approach that job? What do they take from that?

Fradkin: I think before we started this research, we thought one of two things might be going on, or both could be going on at the same time. One is they see that not a lot of people have applied to the job, and they’re like, “Great, I’m going to be one of the first to apply, and it’s going to give me a leg up.” The other thing that we thought might be happening is that maybe they’re like, “Oh, well, not a lot of people apply to this job. Maybe it’s not very good.” Alternatively, let’s say 200-plus people applied to the job already—maybe you’re not going to even try. 

We found the biggest effect was when very few people had applied to the job, people are much more likely to apply. When they see they’d be one of the first to apply, people are a lot more likely to apply.

The Brink: How long can a job be posted before people will start deciding not to apply?

Fradkin: This is actually quite quick. And we have to remember that different jobs have different expectations here. Imagine a company is hiring a CEO—that’s going to be a half-year process, right? So for very skilled, niche, specific jobs, it’s not going to be the case that you have to apply right away. But I think for the types of jobs that were on Jobs on Facebook, a lot of people could do those jobs. In that case, at least, we found that in the first few days you’re more likely to apply to the job than if it was after five days.

The Brink: What can people take from this research and how can they apply it to help them in the job search?

Fradkin: Applying earlier is better. So, monitoring for new postings—you might even have an agent do that for you—and getting ready to apply as soon as a job is posted.

The Brink: How do you think the way people act on job sites is different from how they might respond to an offline posting?

Fradkin: The timeline is different. If you were posting something in a newspaper or a retail storefront, there wouldn’t be the expectation that, if you didn’t apply within a couple of days, you’d be lost in the crowd. But I think online it’s so easy to apply and so easy to see when a job is posted, that everything becomes fast.

The Brink: You previously worked as a data scientist at Airbnb. What did you learn there about the way people use online platforms and markets?

Fradkin: People are very sensitive to even tiny frictions. When Airbnb first started, you had to exchange a lot of messages between the guests and the hosts to get to a transaction. And that communication process broke down very easily. When people experienced that, they were substantially less likely to use the platform. They would go to the hotel sector, because a hotel is very frictionless. As a platform, what Airbnb had to do was incentivize the hosts to create a more frictionless experience for the guest. 

Another aspect that I thought really hard about is the reputation system. When anyone can be a host, how do you know which listings are good or not? And the reputation system is one of those key points—but of course, it’s not perfect. Everyone understands that there are a lot of five-star ratings for listings that are pretty mediocre, so what does a five-star rating mean? 

The Brink: How do you approach selecting an Airbnb rental? Any tips for getting a great stay?

Fradkin: I have over 30 Airbnb stays. I think relative to other folks, I paid very specific attention to the reviews and, in particular, what I’m looking for is a very high average rating. Sometimes people are like, well, it’s a 4.7 and 4.7 could be a good rating. But, actually, given what I know about the distribution of online ratings on Airbnb, I look for higher average ratings. 

The other thing I do is a keyword search in the text of the reviews; I’m very mindful of noise-related issues, because I really care about quality of sleep. I try to just read the text of the bad reviews, because sometimes you can see that this person has unreasonable standards. But especially for listings with not that many reviews, it’s a really informative thing: something like, getting to this place was hard or the shower doesn’t work—you don’t want to stay at an Airbnb where the shower doesn’t work.

The Brink: Is there any more job platform research that you’re interested in doing, or anything you’re working on with online platforms that would be helpful to people in the future?

Fradkin: One set of projects that I’m thinking pretty hard about is: Do we audit digital platforms? I’m sure you’ve seen various lawsuits and laws about big tech platforms. Maybe they’re behaving unfairly, maybe they’re not behaving unfairly. As outsiders, as researchers, how can we measure whether a company is behaving unfairly?

Like everyone else, I’m also thinking a lot about AI and what does that enable for digital platform design? Are we going to have an [AI] agent acting on our behalf, doing our search for us, talking to another agent on the other side and making that transaction happen? And, if we are, how should that work? Is that going to be good? Is that going to be bad? That’s a really fun thing to think about.