College admissions in Australia, by Guillen, Kesten, Kiefer, and Melatos

 Here's a working paper from the University of Sydney that looks at the New South Wales college admissions clearinghouse in which students receive (accurate but unclear) advice from the clearinghouse operator, together with clear but incorrect advice from individual universities.  In an experiment, they look at the effects of these different kinds of advice when presented separately and together.

A Field Evaluation of a Matching Mechanism: University Applicant Behaviour in Australia by Pablo Guillen Onur Kesten, Alexander Kiefer, and Mark Melatos  December 2020

"Abstract: The majority of undergraduate university applications in the state of New South Wales –Australia’s largest state – are processed by a clearinghouse, the Universities Admissions Centre (UAC). Applicants submit an ordered list of degree preferences to UAC which applies a matching algorithm to allocate university places to eligible applicants. The algorithm incorporates the possibility of a type of “early action” through which applicants receive guaranteed enrolments. Applicants receive advice on how to construct their degree preference list from multiple sources (including individual universities). This advice is often confusing, inconsistent with official UAC advice or simply misleading. To evaluate the policy implications of this design choice, we run a large sample (832 observations) experiment with experienced participants in a choice environment that mimics the UAC application process and in which truth telling is a dominant strategy. We vary the advice received across treatments: no advice, UAC advice only, (inaccurate) university advice only, and both UAC and university advice together. Overall, 75.5% of participants fail to use the dominant strategy. High rates of applicant manipulation persist even when applicants are provided with accurate UAC advice. We find that students who attend non-selective government schools are more prone to use strictly dominated strategies than those who attend academically selective government schools and private schools."

The matching algorithm, in which applicants are allowed to list only six choices, is described as follows:

"The algorithm used by UAC sequentially checks each applicant’s eligibility for a degree starting with her first choice. It is therefore reminiscent of the Boston mechanism widely used for school choice in the U.S. (Abdulkadiroglu et al, 2005; Abdulkadiroglu et al, 2006) and college admissions in China (Chen and Kesten, 2017) among other places. However, the absence of formal capacity constraints (on university enrolments) makes this Australian context a unique instance in which the outcome of the algorithm also coincides with that of the celebrated Deferred Acceptance (DA) algorithm of Gale and Shapley (1962). Due to this equivalence, the UAC algorithm does not inherit the strategic vulnerability of the Boston algorithm. Consequently, students are still able to construct their preferred degree list in a manner that is consistent with their true preferences.4

"While the UAC admissions system appears similar to a typical college admissions problem (see, e.g., Roth and Sotomayor, 1991 and Balinski and Sönmez, 1998), universities in NSW can influence student applications through an additional channel. To limit the uncertainty faced by applicants,5 many universities often grant applicants “guaranteed entry” options.6 These schemes represent a university’s commitment to an individualised entry requirement for a particular degree, subject to the candidate’s achievement of a certain score. This innovative feature of the UAC system can be viewed as the centralized or algorithmic embodiment of “early decision” schemes used by over two-thirds of top colleges in the US (see, e.g., Avery, Fairbanks, and Zeckhauser, 2004) that admit students through a decentralized system.7 Indeed, we are not aware of any other centralized college admissions system that has this type of feature. Under the current UAC algorithm, if an applicant includes a guaranteed entry degree in her preference list, this implies that she will not be considered for any degree that she has listed lower on her list provided that she attains the pre-announced entry score."

They conclude in part that

"accurate, albeit somewhat complicated, advice may fail to mitigate the impact of inaccurate (but straightforward) advice."