Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Why Finnish polytechnics reject top applicants

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Finnish polytechnic assignment rejects top applicants unnecessarily, and a counterfactual re-run shows a redesigned mechanism would reassign a quarter of them.

desk verdict 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. read the letter →

arxiv 1908.05443 v1 pith:4RL6Q5FY submitted 2019-08-15 econ.GN q-fin.EC

Kristian Koerselman This is my paper · ORCID
classification econ.GNq-fin.EC
keywords centralizedadmissionsdeferredacceptanceeducationalselectionFinnishpolytechnicsstudentplacementadmissionscoresre-applicationqueuinginhighereducation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

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.

What carries the argument

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.

What would settle it

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.

Watch

Extended reading notes

Core claim

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.

Load-bearing premise

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.

Editorial extensions

If this is right

  • 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.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • 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.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

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.

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 (4)
  1. [Section 4, Data and methods] 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.
  2. [Section 4 / Table 4, row (6)] 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.
  3. [Section 4, Data and methods; Section 5, Results] 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.
  4. [Section 5, Table 4 and Figure 1] 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.
minor comments (5)
  1. [Table 1] 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.
  2. [Section 5, first paragraph] 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.
  3. [Figure 1] 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.
  4. [Table 5] 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.
  5. [References] 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.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: counterfactual reassignments are simulated from observed future applications and mechanical score changes.

full rationale

The paper's central result is generated by feeding observed admission scores and (possibly extended) application lists through a deferred acceptance algorithm and comparing assignments. No parameter is fitted to the target outcome: the counterfactual applications are taken from applicants' own 2012/13 clearinghouse records, the score counterfactuals are mechanical (removal of first-choice points, first exam made valid), and the quality outcome is the field-specific matriculation GPA rank, chosen specifically because it is not an input to the score changes. The descriptive statistics in Tables 2-3 and the stable-assignment replication do not presuppose the conclusion. The Table 5 linear probability models are ancillary and are explicitly not used to infer behavior under counterfactual rules. Self-citations (Koerselman & Uusitalo 2014; Hämäläinen et al. 2017) appear only as background context, not as load-bearing premises. The main identifying assumption that 2012/13 applications reveal 2011 acceptability conditional on rejection is a substantive assumption that could overstate the counterfactual if preferences changed, but an assumption about external validity is not a circular reduction of the derivation to its inputs.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper's central claim rests on domain assumptions about how to interpret later-year applications as evidence of earlier acceptability, how to proxy missing administrative identifiers, and how to measure applicant quality. These are stated transparently but are not independently verified. No free parameters are fit, and no new entities are postulated.

assumptions (5)
  • domain assumption Programs acceptable to an applicant in 2012 and 2013 are also acceptable to that applicant in 2011, conditional on rejection from programs applied to in 2011.
    This assumption turns later-year applications into counterfactual 2011 applications and is the basis for Table 4 rows 2, 4, and 6; if false, the unnecessary rejection counterfactual overstates the mechanism's role.
  • domain assumption Combinations of polytechnic name and program name, plus the number of simultaneous offers, are adequate proxies for true program identifiers and quota.
    The authors lacked true identifiers and quota; they replicate 98% of decisions, but the remaining 2% discrepancy could affect counterfactual reassignment counts.
  • domain assumption Field-specific matriculation GPA rank is a valid measure of applicant quality for comparing assignments.
    Used as the outcome variable because it is not directly changed by counterfactual score adjustments; however, matriculation GPA carries only 0.28 weight in the admission score, so it captures a partial aspect of quality.
  • domain assumption The replicated program-proposing deferred acceptance assignment is a valid benchmark.
    All counterfactuals are compared to the replicated benchmark rather than the actual assignment; 98% agreement supports this, but the 2% gap is unmodeled.
  • standard math Deferred acceptance algorithm produces stable matchings; if applicant- and program-proposing DA coincide, the stable matching is unique.
    The paper relies on this property to conclude there is only one stable assignment conditional on applications; this is a standard result cited to Gale-Shapley and Roth-Peranson.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Why Finnish polytechnics reject top applicants." pith.science (2026). https://pith.science/paper/4RL6Q5FY

@misc{pith2026190805443,
  author       = {Pith},
  title        = {Pith review of: Why Finnish polytechnics reject top applicants},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4RL6Q5FY}},
  note         = {Machine review of arXiv:1908.05443}
}
read the original abstract

I use a panel of higher education clearinghouse data to study the centralized assignment of applicants to Finnish polytechnics. I show that on a yearly basis, large numbers of top applicants unnecessarily remain unassigned to any program. There are programs which rejected applicants would find acceptable, but the assignment mechanism both discourages applicants from applying, and stops programs from admitting those who do. A mechanism which would admit each year's most eligible applicants has the potential to substantially reduce re-applications, thereby shortening the long queues into Finnish higher education.

Figures

Figures reproduced from arXiv: 1908.05443 by the authors.

Figure 1
Figure 1. The grade rank distribution of assigned applicants. [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 33 canonical work pages

  1. [1]

    , author Che, Y.-K

    author Abdulkadiro g lu, A. , author Che, Y.-K. , & author Yasuda, Y. ( year 2011 ). Resolving Conflicting Preferences in School Choice: The ``Boston Mechanism'' Reconsidered. journal American Economic Review \/ , volume 101 \/ , pages 399--410

    work page 2011

  2. [2]

    , author Pathak, P

    author Abdulkadiro g lu, A. , author Pathak, P. A. , & author Roth, A. E. ( year 2005 a ). The New York City High School Match. journal American Economic Review \/ , volume 95 \/ , pages 364--367

    work page 2005

  3. [3]

    , author Pathak, P

    author Abdulkadiro g lu, A. , author Pathak, P. A. , & author Roth, A. E. ( year 2009 ). Strategy-proofness versus Efficiency in Matching with Indifferences: Redesigning the NYC High School Match. journal American Economic Review \/ , volume 99 \/ , pages 1954--78

    work page 2009

  4. [4]

    , author Pathak, P

    author Abdulkadiro g lu, A. , author Pathak, P. A. , author Roth, A. E. , & author S \"o nmez, T. ( year 2005 b ). The Boston Public School Match. journal American Economic Review \/ , volume 95 \/ , pages 368--371

    work page 2005

  5. [5]

    , & author S\" o nmez, T

    author Abdulkadiroğlu, A. , & author S\" o nmez, T. ( year 2003 ). School Choice: A Mechanism Design Approach . journal American Economic Review \/ , volume 93 \/ , pages 729--747

    work page 2003

  6. [6]

    , author Kanoria, Y

    author Ashlagi, I. , author Kanoria, Y. , & author Leshno, J. D. ( year 2017 ). Unbalanced Random Matching Markets: The Stark Effect of Competition. journal Journal of Political Economy \/ , volume 125 \/ , pages 69--98

    work page 2017

  7. [7]

    author Azevedo, E. M. , & author Leshno, J. D. ( year 2016 ). A Supply and Demand Framework for Two-Sided Matching Markets. journal Journal of Political Economy \/ , volume 124 \/ , pages 1235--1268

    work page 2016

  8. [8]

    , & author S \"o nmez, T

    author Balinski, M. , & author S \"o nmez, T. ( year 1999 ). A Tale of Two Mechanisms: Student Placement. journal Journal of Economic theory \/ , volume 84 \/ , pages 73--94

    work page 1999

Show all 34 references
  1. [9]

    , author Magnac, T

    author Carvalho, J.-R. , author Magnac, T. , & author Xiong, Q. ( year 2018 ). College Choice, Selection and Allocation Mechanisms: A Structural Empirical Analysis

    2018

  2. [10]

    , & author Pereyra, J

    author Chen, L. , & author Pereyra, J. S. ( year 2019 ). Self-selection in School Choice. journal Games and Economic Behavior \/ , volume 117 \/ , pages 59 -- 81

    2019

  3. [11]

    , & author Askling, B

    author Cliffordson, C. , & author Askling, B. ( year 2006 ). Different Grounds for Admission: Its Effects on Recruitment and Achievement in Medical Education. journal Scandinavian Journal of Educational Research \/ , volume 50 \/ , pages 45--62

    2006

  4. [12]

    , author Cohn, S

    author Cohn, E. , author Cohn, S. , author Balch, D. C. , & author Bradley Jr, J. ( year 2004 ). Determinants of Undergraduate GPAs: Sat Scores, High-School GPA and High-School Rank. journal Economics of Education Review \/ , volume 23 \/ , pages 577--586

    2004

  5. [13]

    author Dillon, E. W. , & author Smith, J. A. ( year 2017 ). Determinants of the Match Between Student Ability and College Quality. journal Journal of Labor Economics \/ , volume 35 \/ , pages 45--66

    2017

  6. [14]

    , author Kominers, S

    author Dur, U. , author Kominers, S. D. , author Pathak, P. A. , & author Sönmez, T. ( year 2018 ). Reserve Design: Unintended Consequences and the Demise of Boston’s Walk Zones . journal Journal of Political Economy \/ , volume 126 \/ , pages 2457--2479

    2018

  7. [15]

    , author Libassi, C

    author Dynarski, S. , author Libassi, C. , author Michelmore, K. , & author Owen, S. ( year 2018 ). Closing the Gap: The Effect of a Targeted, Tuition-Free Promise on College Choices of High-Achieving, Low-Income Students. note National Bureau of Economic Research Working Paper 25349

    2018

  8. [16]

    , & author Yenmez, M

    author Echenique, F. , & author Yenmez, M. B. ( year 2015 ). How to Control Controlled School Choice. journal American Economic Review \/ , volume 105 \/ , pages 2679--94

    2015

  9. [17]

    , author Grenet, J

    author Fack, G. , author Grenet, J. , & author He, Y. ( year 2018 ). Beyond Truth-Telling: Preference Estimation with Centralized School Choice. note American Economic Review, forthcoming

    2018

  10. [18]

    , & author Shapley, L

    author Gale, D. , & author Shapley, L. S. ( year 1962 ). College Admissions and the Stability of Marriage . journal American Mathematical Monthly \/ , volume 69 \/ , pages 9--15

    1962

  11. [19]

    a m \"a l \

    author H \"a m \"a l \"a inen, U. , author Koerselman, K. , & author Uusitalo, R. ( year 2017 ). title Graduation Incentives Through Conditional Student Loan Forgiveness \/ . type Technical Report IZA Discussion Papers

    2017

  12. [20]

    , & author Samet, D

    author Holzman, R. , & author Samet, D. ( year 2014 ). Matching of Like Rank and the Size of the Core in the Marriage Problem. journal Games and Economic Behavior \/ , volume 88 \/ , pages 277--285

    2014

  13. [21]

    author Hoxby, C. M. , & author Avery, C. ( year 2012 ). title The Missing ``One-Offs'': The Hidden Supply of High-Achieving, Low Income Students \/ . type Technical Report National Bureau of Economic Research

    2012

  14. [22]

    author Hoxby, C. M. , & author Turner, S. ( year 2015 ). What High-Achieving Low-Income Students Know About College. journal American Economic Review \/ , volume 105 \/ , pages 514--17

    2015

  15. [23]

    , & author Mahdian, M

    author Immorlica, N. , & author Mahdian, M. ( year 2005 ). Marriage, Honesty, and Stability. In booktitle Proceedings of the sixteenth annual ACM-SIAM symposium on Discrete algorithms \/ (pp. pages 53--62 ). organization Society for Industrial and Applied Mathematics

    2005

  16. [24]

    ( year 2010 )

    author Kesten, O. ( year 2010 ). School Choice with Consent. journal The Quarterly Journal of Economics \/ , volume 125 \/ , pages 1297--1348

    2010

  17. [25]

    , & author Uusitalo, R

    author Koerselman, K. , & author Uusitalo, R. ( year 2014 ). The Risk and Return of Human Capital Investments. journal Labour Economics \/ , volume 30 \/ , pages 154--163

    2014

  18. [26]

    , & author Pathak, P

    author Kojima, F. , & author Pathak, P. A. ( year 2009 ). Incentives and Stability in Large Two-Sided Matching Markets. journal American Economic Review \/ , volume 99 \/ , pages 608--27

    2009

  19. [27]

    , & author McNally, S

    author Machin, S. , & author McNally, S. ( year 2005 ). Gender and Student Achievement in English Schools. journal Oxford Review of Economic Policy \/ , volume 21 \/ , pages 357--372

    2005

  20. [28]

    title Education at a Glance 2014: OECD Indicators \/

    author OECD ( year 2014 ). title Education at a Glance 2014: OECD Indicators \/ . type Technical Report

    2014

  21. [29]

    author Pathak, P. A. ( year 2017 ). What Really Matters in Designing School Choice Mechanisms. In editor B. Honoré , editor A. Pakes , editor M. Piazzesi , & editor L. Samuelson (Eds.), booktitle Advances in Economics and Econometrics: Eleventh World Congress \/ (p. pages 176–...

    2017

  22. [30]

    ( year 1989 )

    author Pittel, B. ( year 1989 ). The Average Number of Stable Matchings. journal SIAM Journal on Discrete Mathematics \/ , volume 2 \/ , pages 530--549

    1989

  23. [31]

    author Roth, A. E. , & author Peranson, E. ( year 1999 ). The Redesign of the Matching Market for American Physicians: Some Engineering Aspects of Economic Design. journal American Economic Review \/ , volume 89 \/ , pages 748--780

    1999

  24. [32]

    o nmez, T. , & author \

    author S \"o nmez, T. , & author \"U nver, M. U. ( year 2011 ). Matching, Allocation, and Exchange of Discrete Resources. In booktitle Handbook of Social Economics \/ (pp. pages 781--852 ). publisher Elsevier volume volume 1

    2011

  25. [33]

    , & author Zhong, X

    author Wu, B. , & author Zhong, X. ( year 2014 ). Matching Mechanisms and Matching Quality: Evidence from a Top University in China. journal Games and Economic Behavior \/ , volume 84 \/ , pages 196--215

    2014

  26. [34]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.