MARKET FOR RESIDENT ASSISTANTS

THE TL;DR: utilizing auction methodology might be the key to maximizing satisfaction simultaneous mass hiring scenarios.

The Problem

In January of 2015, I was selected to be a Community Advisor at Carnegie Mellon. Community Advisors act as Head Resident Assistants: managing the daily internal logistics of a dormitory; overseeing a staff of resident assistants; responding to high level emergencies.  The first task of a Community Advisor, immediately after hiring, is selecting a staff in tandem with your Housefellow. After participating in the hiring process from the side of the employer, I noticed some inefficiencies within the matching system.

Unsure of how to provide each candidate equal consideration, the process allowed each residential team an opportunity to ask to hire each candidate from a shared pool. This created heavy time requirements due to the volume of applications, leaving residential teams to negotiate the environment prior in an effort to ease navigation through the process with fewer frictions. I found this entire markedly similar to the original market for hiring law clerks, and noticing signs of market unraveling, showing that this system was not working as an efficient matching system on both sides, Additionally, because each resident assistant is allocated by how vocal a resident assistant team is, more timid individuals were less likely to advocate for their needs (or perhaps less vocal individuals didn't care as much, but it couldn't be definitively shown). I figured I might take a gander at using my coursework to see if there's a solution.

Why the Hiring Process Isn't a Matching Problem

Upon first glance, it seemed the process simply needed an implementation of the deferred acceptance algorithm, the solution to the "stable marriage problem", which you can learn about with this cool video here. It turns out that doesn't actually work for a couple reasons. 

First, matching problems mean that both sides know exactly what they want and they can put together a definitive preference ranking. However, many new applicants into the RA selection process don't have a meaningful understanding of the requirements and values embodied in each house community. They're unable to put together a fully informed ranking of preferred residencies of employment.

Second, the hiring process for resident assistants requires diversity in staff selection. However, with the way that matching algorithms work, you have little control over the numerous extra factors one must consider outside of gender alone: nationality, religion, race, socioeconomic status, international student status (just to name a few). Any effort to maximize diversity would be incredibly difficult to implement in the traditional style of quotas. Additionally, with the way deferred acceptance works, while one candidate may temporarily be on one's staff (and other candidates placed to fill any gaps in diversity), if a better offer in given in terms of that candidate's ranked preferences, they disappear, leaving a very specifically shaped hole in their stead.

These two problems add up to a large headache when trying to implement a centralized matching system. This is because, in reality, the initial process of hiring resident assistants is actually one large auction.

The Simultaneous Ascending Auction

Most auctions that people are familiar with involve bidding on one item at a time, with a person with a gavel yelling in the front of the room. But what about if you're trying to auction 100 things? Or 1000 things? And there are fewer buyers than there are things? 

The simultaneous ascending auction popped up as a solution to the FCC's need to sell ownership of several radio channels to a few large radio conglomerates and several smaller radio stations. The auction style allows everyone to bid on as many items as they want; the only contingency is if two buyers want the same item, the price will go up. This bidding will continue until all items have only one bidder. 

The simultaneous ascending auction is cool because it solves a couple problems. There's no incentive to lay claims to an item early on, because the bidder who wants an item the most will be able to buy it. Additionally, if a bidder loses an item early on, they're able to take the L and change their bidding strategy without much harm being done.

This has really cool implications for mass hiring practices from a singular pool. First, a person more timid does not have to challenge the room. Instead, they can talk with their money. Second, a person who has built an can build their entire team after securing their best player, knowing that they need to conserve their resources for the rest of the team as well. Lastly, an operation more specific to resident assistant selection, the bidding on candidates by multiple teams provides for an objective assessment of merit, a problem often critiqued in any setting where students are hiring other students.

Theoretically, this solves a lot of problems plaguing resident assistant selection. Of course, there are always difficulties that emerge with testing. You can read more about the theory of it in my term paper.

Implementation

After continuously bothering a fellow CA, Kevin Karol, to help me figure out implementation, Kevin was great enough to think through a way to try beta testing through Google Sheets. Because I am nothing if not persistent, I asked the Residential Education office if I could run a pilot prior to the actual selection process.

Each residential team was assigned 10 points to allocate per each empty staff space they needed to fill. The bidding happened all at once, until each spot on every residential team was filled. The impressive part was that, even with longer breaks for talking, 50 resident assistants were assigned to virtual dormitories (with fewer spots available) within an hour, a stark contrast to the normal eight hours of resident assistant selection that occurs with around 200 candidates. However, there were some issues present as well. 

The single gender dorms faced a marked disadvantage with a restriction on candidates available for bid. With the only single gender dorm on campus being an all-male dorm, the male candidates had higher competition between residential teams for selection than the female candidates did. There could be a way to appropriately account for this disadvantage by scaling the points available to the all male dormitory by the size of the additional competition they have to face. For instance, if there are 11 teams, then the male dormitory may receive 10 points per candidate while the rest of the teams receive 9 points. I'm not completely sure this would work, however, and would be interested to see how this advantage plays in practice.

While the system was never fully implemented, you can read about it on the Tepper School of Business website. The entire process was enjoyable and was extremely rewarding. I'm so thankful to CMU for giving me the opportunity to run this trial with them, which I cannot say would happen at most other colleges. One of the things I am most thankful for is my school's adherence to one of the major rules of improvisation, "Yes, and..."