{"id":"54505943-6555-442e-9bcf-36b9122aff9b","arxiv_id":"1908.05443","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The Finnish polytechnic assignment mechanism unnecessarily rejects many top applicants because of first-choice bonuses and entrance exam rules, and a counterfactual using later applications shows admission quality would improve if the mechanism changed.","lead":"Using Finnish clearinghouse data, the paper shows that the centralized polytechnic assignment mechanism leaves many high-scoring applicants unassigned each year, even when acceptable programs could have admitted them. A generalist reader should care because the same mechanism design flaw may explain long queues and delayed graduation in Finnish higher education.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The main counterfactual assumes 2012/13 applications reveal 2011 acceptability; the largest gains also import future exam scores, so the 'unnecessary rejection' headline may overstate mechanism failure.","rationale":"The descriptive finding that 34% of top-third applicants by admission score remain unassigned is robust and does not depend on the counterfactual; this is genuine evidence that the mechanism leaves many high-scoring applicants without seats. The counterfactual is also clearly labeled as resting on an assumption about future applications. However, the paper's headline-strength numbers, 24% reassigned and +4.59 percentile-rank gain, come from a bundle of three changes: extended applications, removal of first-choice points, and cross-year portability of entrance exams. The first depends on the untested acceptability assumption, and the third imports future scores into the past. Rows (3) and (5) of Table 4 show that score design alone has meaningful effects even without future applications, which supports part of the mechanism-failure story. The weakest link is therefore the joint counterfactual, not the descriptive claim. A conservative re-estimation that restricts future applications to same-field programs for unassigned applicants and removes future exam scores would show whether the strong headline survives. This is an addressable robustness concern rather than a fatal flaw, so the conditional verdict stands.","tokens_in":13088,"tokens_out":8808,"duration_ms":89094,"concrete_test":"Recompute Table 4 row (6) under a restricted acceptability definition: include a 2012/13 application in the 2011 counterfactual only if (i) the applicant was unassigned in 2011, and (ii) the later program lies in the same field as one of the applicant's 2011 applications; simultaneously drop the cross-year entrance-exam portability and keep 2011 scores only. If the reassignment share falls from 24% to near the 15% of row (4) or below, the 'unnecessarily rejected' claim is not robust to the acceptability assumption; if it remains above 20%, the concern is minor.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's bridge from 'top applicants are unassigned' to 'they are unnecessarily rejected' is Section 4's assumption that programs applied to in 2012/13 would also have been acceptable in 2011 conditional on 2011 rejection. This is untested. Rejected applicants' later applications may reflect changed circumstances, new information about preferences, or willingness to accept programs they would not have chosen in 2011. If so, adding those applications to the 2011 choice set assigns applicants to programs they did not want then, inflating both the 24% reassignment and the +4.59 GPA-rank gain in Table 4 row (6). The same row also makes the first entrance exam taken in each field valid for all 2011 applications in that field, effectively importing scores from 2012/13 into 2011. This changes the information and score environment rather than isolating the mechanism's rejection of 2011 top applicants. Since rows (3) and (5) show score-only changes alone reassign 5-14% of applicants, the headline 'unnecessary rejection' is only partly identified without the acceptability assumption.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses Finnish clearinghouse data covering all polytechnic applications in 2011-2013 to argue that the centralized assignment mechanism unnecessarily rejects top applicants. Descriptive statistics show that 54% of applicants in the top third of their applied programs by matriculation GPA and 34% by actual admission score remain unassigned. The author then simulates counterfactual assignments by extending application lists with later-year applications, removing first-choice bonus points, and making entrance exams valid across years; the most comprehensive counterfactual reassigns 24% of applicants and raises the mean field-specific matriculation GPA rank of admitted applicants by 4.59 percentiles. The paper also reports that the program-proposing and applicant-proposing stable assignments coincide conditional on submitted applications, and that lower-ranked applications and entrance exam participation predict later non-acceptance and re-application.","tokens_in":13300,"tokens_out":4524,"duration_ms":45365,"significance":"If the counterfactual results are taken at face value, the paper makes a useful contribution by showing that a stable centralized mechanism can still attenuate selection when application lists are short, priorities include first-choice bonuses, and entrance exams are repeated across years. The descriptive facts about top applicants being unassigned are clean, and the 98% replication of assignment decisions is a credible benchmark. The paper is also transparent about its proxy limitations and about the fact that the outcome measure is chosen to be unaffected by the counterfactual score changes. The main weakness is that the headline 'unnecessary rejection' conclusion is only as strong as the untestable acceptability assumption in Section 4, so the paper needs a substantial revision to either test or bound that assumption and to clarify which parts of the counterfactual gains come from mechanism design as opposed to importing future information.","major_comments":[{"comment":"The counterfactual analysis rests on the assumption that programs applied to in 2012 or 2013 would have been acceptable to the applicant in 2011 conditional on rejection from the 2011 applications. This assumption is stated but not tested, and later applications may reflect changed circumstances, new information, or changed preferences rather than stable acceptability. Because Table 4 rows (2), (4), and (6) all use extended applications, the 24% reassignment and +4.59 percentile rank in row (6) are not identified without this assumption. The paper should add a sensitivity analysis or bounds that do not require full acceptability, for example by adding future applications only for applicants who were rejected in 2011 from the same field, or by varying the fraction of future applications treated as acceptable.","section":"Section 4, Data and methods"},{"comment":"The counterfactual that makes the first entrance exam taken in each field valid for all applications in that field imports exam scores from 2012 and 2013 into the 2011 application round. This changes the information and score environment rather than isolating the mechanism's rejection of 2011 top applicants. The paper itself shows in rows (3) and (5) that score-only changes reassign 5% and 14% of applicants, so the headline row (6) combines the acceptability assumption with these score changes. The discussion should explicitly decompose the contribution of each component and should not attribute the full +4.59 percentile gain to the mechanism's handling of 2011 applicants.","section":"Section 4 / Table 4, row (6)"},{"comment":"Program identifiers and quota are proxied by combinations of polytechnic name and program name and by the number of simultaneous offers, and the counterfactual comparisons use the replicated assignment as the benchmark while descriptive statistics use the actual assignment. The 98% replication rate is reassuring, but the paper does not report how the main counterfactual numbers change when the 2% of non-replicated decisions are treated differently, nor whether the proxy for quota is likely to bias the counterfactual assignments in a particular direction. A robustness check that recomputes Table 4 under an alternative handling of the non-replicated decisions would strengthen the paper's central claim.","section":"Section 4, Data and methods; Section 5, Results"},{"comment":"The headline quality improvement is measured only by field-specific matriculation GPA rank, a component with effective weight 0.28 in the admission score as shown in Table 1. The paper's argument that the mechanism attenuates stated selection criteria would be stronger if it also reported corresponding changes in the actual admission score or in other score components, at least descriptively. As written, the +4.59 percentile rank can be driven entirely by the component that the counterfactual changes least, and it does not directly establish that admitted applicants are more suitable under the full stated criteria.","section":"Section 5, Table 4 and Figure 1"}],"minor_comments":[{"comment":"The note says the table shows 'standardized square roots of the variance components,' but this is not standard terminology; please clarify how the effective weights are computed and whether they are intended to sum to one.","section":"Table 1"},{"comment":"The statement that 'not a single applicant is assigned to a different program' under the applicant-proposing assignment should be explicitly conditioned on the replicated application data and realized scores, since the benchmark itself is an approximation of the actual algorithm.","section":"Section 5, first paragraph"},{"comment":"The panels in Figure 1 are difficult to read without a legend; please add explicit identification on the figure itself of which counterfactual assignment corresponds to which panel, rather than relying only on the panel numbers in the caption.","section":"Figure 1"},{"comment":"The table reports coefficients and standard errors but no R-squared or number of clusters; for a linear probability model with many interacted controls, readers would benefit from knowing the model fit and the standard error adjustment.","section":"Table 5"},{"comment":"Several cited works are working papers or forthcoming; please update Carvalho, Magnac, and Xiong, and Fack, Grenet, and He to their published versions if they have appeared by the time of revision.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper's policy conclusion is interesting and the data work is careful, but the central counterfactual rests on a strong, untested assumption about the stability of preferences over time. I would send it back for revision rather than reject: the assumption may not be fully testable, but the paper needs to reframe the headline claims, add sensitivity analyses, and clearly separate the contributions of extended applications, first-choice points, and future entrance exam scores. The scope is appropriate for an applied economics or education policy venue, though the current version is more a policy analysis than a methodological contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, read this one if you care about centralized admissions. The paper's real contribution is a clean, full-system look at Finland's polytechnic assignment: 34% of applicants in the top third of their applied programs by actual admission score end up with nothing. That descriptive fact is robust and well worth knowing.\n\nWhat's actually new: the panel-based counterfactual that adds later-year applications into the 2011 choice set, and the finding that conditional on submitted applications there is a unique stable assignment. The paper is also honest about its data: program IDs and quotas are proxied, and the 98% replication is decent. The outcome measure (field-specific matriculation GPA rank) is sensibly chosen to be unaffected by the score changes they simulate.\n\nThe soft spot is exactly where the stress-test puts it. Section 4 assumes programs applied to in 2012/13 would also have been acceptable in 2011 conditional on rejection from the 2011 list. That's untestable, and later applications may reflect changed circumstances or new information rather than stable tastes. Row (6) of Table 4 also makes the first entrance exam taken in a field valid for all 2011 applications in that field, which imports future scores into the earlier market. So the 24% reassignment and +4.59 GPA rank gain are an upper bound on what mechanism redesign alone would deliver. The paper actually says 'under the assumption' clearly enough, so it's not hiding the problem, but the abstract's claim about reducing re-applications goes beyond what any row simulates.\n\nNone of this is fatal. The descriptive core stands, the counterfactual is clearly labeled as conditional, and the citation pattern is appropriate. What you get is a useful, policy-relevant case study rather than a fundamental mechanism-design insight.\n\nWorth a serious referee. A good referee should push for uncertainty quantification around Table 4 and a more careful interpretation of the re-application claim, but the paper deserves peer review rather than desk rejection. I'd bring it to a reading group if you work on admissions; otherwise, skim the tables and move on.","headline":"Clean descriptive result that Finland's mechanism leaves many top applicants unassigned, but the headline counterfactual overstates mechanism failure by importing future preferences and exam scores.","tokens_in":13769,"tokens_out":1993,"would_cite":true,"duration_ms":19575,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Finnish polytechnic assignment rejects top applicants unnecessarily, and a counterfactual re-run shows a redesigned mechanism would reassign a quarter of them.","keywords":["centralized admissions","deferred acceptance","educational selection","Finnish polytechnics","student placement","admission scores","re-application","queuing in higher education"],"falsifier":"A direct survey of rejected 2011 applicants asking whether they would have accepted a seat in a program they first applied to only in 2012 or 2013 could settle the counterfactual. If a large share said no, the 24 percent reassignment figure overstates the number of unnecessarily rejected applicants; if a large share said yes, the paper's mechanism-failure account is supported. A narrower computational check would compare the replicated deferred acceptance assignment to the actual assignment once true program identifiers and quotas are available, since the 2 percent discrepancy could hide systematic rejection of top applicants.","tokens_in":12903,"feed_emoji":"🎓","tokens_out":7350,"duration_ms":63971,"temperature":0.7,"pith_summary":"The paper argues that the centralized Finnish polytechnic assignment system fails at its own selection criterion: 34 percent of applicants who rank in the top third of their applied programs by the actual admission score end up unassigned to any program. The cause, it claims, is the mechanism rather than applicant preferences: a four-program cap, bonus points for the first-listed choice, and entrance exams that do not transfer across years discourage applications and distort priorities. Using later-year applications as evidence of which programs rejected applicants find acceptable, the author constructs counterfactual assignments that reassign 24 percent of applicants and raise the mean field-specific matriculation GPA rank of admitted applicants by 4.59 percentiles. If this is right, the long queues into Finnish higher education are in part a mechanical artifact, not a necessary consequence of scarce seats or weak demand.","feed_headline":"A third of top applicants end up without a seat in Finnish polytechnics","feed_subtitle":"Dropping first-choice bonuses and reusing entrance exams across years would reassign 24% of applicants and shorten queues.","key_machinery":"The central object is the Finnish polytechnic clearinghouse, a centralized assignment run by a program-proposing deferred acceptance algorithm over program-specific admission scores built from matriculation exam GPA, entrance exam results, and a bonus for listing a program first. The argument's engine is a counterfactual reconstruction of the 2011 assignment: the author appends each applicant's 2012 and 2013 applications to the bottom of the 2011 list, removes the first-choice bonus, and makes the first entrance exam taken in a field valid for all applications in that field, then re-runs the algorithm. This lets the paper decompose the mechanism's effect into the contribution of limited applications, of first-choice points, and of year-specific entrance exams, while using field-specific GPA rank as an outcome measure that is not itself altered by the counterfactual score changes.","core_discovery":"On the paper's own terms, the central discovery is that the Finnish polytechnic assignment is stable but not selective. Conditional on the applications actually submitted and the realized admission scores, there is a unique stable assignment—the program-proposing and applicant-proposing deferred acceptance algorithms coincide—so the mechanism never violates priorities within the submitted lists. Yet the submitted lists themselves are short and distorted: applicants average 2.77 programs, only 19,048 of 50,894 applicants use the maximum of four, and admission to a non-first choice is rare. Re-running the assignment with future applications added, first-choice bonus points removed, and the first entrance exam made valid for all applications in the same field reassigns 24 percent of applicants and improves the field-specific matriculation GPA rank of admitted applicants by 4.59 percentiles, with most of the gain coming from the top fifth of the grade distribution. The mechanism is therefore attenuating the stated selection criteria and creating an incentive to re-apply, which the paper identifies as a source of long queues and high graduation ages.","pith_inferences":["The paper's outcome measure, field-specific GPA rank, is only a proxy for applicant quality; if entrance exams capture program-specific aptitude, the true selection loss from discarding exam scores across years could differ from the 4.59 percentile estimate.","The 2 percent replication gap between the actual and replicated assignments suggests the real mechanism may be even more selective on hidden criteria, so the counterfactual results should be read as lower bounds on the reassignment margin.","If the mechanism were changed to admit the most eligible applicants, policymakers could test the queuing hypothesis directly by observing whether the number of re-applicants falls in subsequent admission rounds.","The same panel method could be applied to Finnish university admissions, which are largely decentralized, to ask whether centralized assignment is the active ingredient behind the attenuation."],"forward_implications":["Removing the first-choice bonus alone would reassign 5 percent of applicants and improve the mean matriculation GPA rank of admitted applicants by 0.56 percentile.","Removing the bonus and making entrance exams valid across years would reassign 14 percent of applicants and improve the grade rank by 3.97 percentiles.","Adding future applications as well reassigns 24 percent of applicants and improves the grade rank by 4.59 percentiles, concentrated in the top fifth of the grade distribution.","Admitted applicants who were placed at a second-, third-, or fourth-listed program are 10 to 20 percentage points more likely to re-enter the application system in a later year, so the mechanism generates re-application queues.","A mechanism that assigned each year's most eligible applicants would shorten the queues into Finnish higher education and could lower the average graduation age."],"supporting_citations":[{"why":"Defines deferred acceptance and stability, the benchmark mechanism the paper uses to characterize the Finnish assignment.","marker":"Gale & Shapley, 1962"},{"why":"Establishes the student-optimal stable mechanism and the school-choice perspective that the paper contrasts with selection on applicant quality.","marker":"Abdulkadiroğlu & Sönmez, 2003"},{"why":"Provides the result that the set of stable assignments is typically small in large markets, which supports the paper's uniqueness finding.","marker":"Roth & Peranson, 1999"},{"why":"Shows that limited applications tend to make the set of stable assignments small, supporting the uniqueness result the paper relies on.","marker":"Immorlica & Mahdian, 2005"},{"why":"Shows how a first-choice-driven mechanism creates all-or-nothing application dynamics under uncertainty, the mechanism channel the paper blames for short application lists.","marker":"Abdulkadiroğlu et al., 2011"},{"why":"Supplies the idea that applicants may apply to exactly the most preferred programs that would accept them, which motivates using later-year applications as revealed acceptability.","marker":"Fack et al., 2018"},{"why":"Demonstrates counterfactual selection improvements when applicants can apply to more programs, a direct template for the paper's extended-application counterfactuals.","marker":"Carvalho et al., 2018"},{"why":"Documents that misperceived admission chances keep top applicants from applying, supporting the paper's proposal to reveal admission probabilities before application.","marker":"Chen & Pereyra, 2019"},{"why":"Provides empirical evidence that assignment mechanisms attenuate selection on application scores, the effect the paper measures for Finnish polytechnics.","marker":"Wu & Zhong, 2014"}],"fun_headline_variants":["Top applicants left out by Finnish polytechnics' stable matching","Stable but unfair: Finnish polytechnic assignment","Finnish polytechnics: stable matching, lost top applicants","How Finland's assignment mechanism loses top applicants"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole counterfactual hinges on treating the programs an applicant applied to in 2012 or 2013 as programs that would also have been acceptable in 2011, conditional on rejection from the programs applied to in 2011; if later applications reflect changed circumstances or new information rather than stable preferences, the reassignment numbers are overstated.","fun_headline_variants_meta":{"raw":{"variants":["Top applicants left out by Finnish polytechnics' stable matching","Stable but unfair: Finnish polytechnic assignment","Finnish polytechnics: stable matching, lost top applicants","How Finland's assignment mechanism loses top applicants"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000261,"raw_usage":{"total_tokens":1540,"prompt_tokens":837,"completion_tokens":703,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":453,"completion_tokens_details":{"reasoning_tokens":638}},"tokens_in":453,"tokens_out":703,"duration_ms":7359,"temperature":1.0,"reasoning_tokens":638,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:13:08.461056+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct survey of rejected 2011 applicants asking whether they would have accepted a seat in a program they first applied to only in 2012 or 2013 could settle the counterfactual. If a large share said no, the 24 percent reassignment figure overstates the number of unnecessarily rejected applicants; if a large share said yes, the paper's mechanism-failure account is supported. A narrower computational check would compare the replicated deferred acceptance assignment to the actual assignment once true program identifiers and quotas are available, since the 2 percent discrepancy could hide systematic rejection of top applicants.","supporting_citations":[{"cited_title":", & author Shapley, L","cited_arxiv_id":null,"evidence_quote":"Defines deferred acceptance and stability, the benchmark mechanism the paper uses to characterize the Finnish assignment."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the result that the set of stable assignments is typically small in large markets, which supports the paper's uniqueness finding."},{"cited_title":", & author Mahdian, M","cited_arxiv_id":null,"evidence_quote":"Shows that limited applications tend to make the set of stable assignments small, supporting the uniqueness result the paper relies on."},{"cited_title":", author Grenet, J","cited_arxiv_id":null,"evidence_quote":"Supplies the idea that applicants may apply to exactly the most preferred programs that would accept them, which motivates using later-year applications as revealed acceptability."},{"cited_title":", author Magnac, T","cited_arxiv_id":null,"evidence_quote":"Demonstrates counterfactual selection improvements when applicants can apply to more programs, a direct template for the paper's extended-application counterfactuals."},{"cited_title":", & author Pereyra, J","cited_arxiv_id":null,"evidence_quote":"Documents that misperceived admission chances keep top applicants from applying, supporting the paper's proposal to reveal admission probabilities before application."}],"review_version":1}