The Arc 2 Closing Move: Commitment 5b Is Dead on Arrival (16-SMD-0-PR), R19's Headline Attenuates 49% to -0.77 bills/month, and the Project's Methodological Signature Crystallizes Into Four Honest Retreats - A Final-Round Review of Scout R22 and Analyst R22

1. Overall Assessment

Two-sentence summary: Analyst's R22 verification that all 16 cohort runners held district seats at exit term and none held PR seats kills Scout R21-R22's Commitment 5b as a within-cohort falsification test but converts the null into a pipeline-selection finding that is more interesting than the moderator test would have been. The NEC-proxy re-windowing attenuates R19's headline from -1.5 to -0.77 bills/month (sign stable, magnitude -49%), which is the project's fourth honest retreat and the single most credibility-enhancing move in the arc because it pre-empts the exact specification-search criticism an observational-design skeptic would raise.

2. Citation Verification (C9)

Crossref-verified this round:

• Carey and Shugart (1995) doi:10.1016/0261-3794(94)00035-2 - CONFIRMED by Crossref: title "Incentives to cultivate a personal vote: A rank ordering of electoral formulas", authors Carey and Shugart, year 1995. Scout R22's framing of the paper as the canonical origin anchor is accurate.
• Hansen (2026) doi:10.1111/lsq.70056 - previously flagged R22 Scout; not re-verified this round because the paper is listed as forthcoming in LSQ and Crossref coverage of forthcoming articles is inconsistent. Flagging for R23+ re-verification when the issue is finalized.
• Im (2025) doi:10.30992/kpsr.2025.3.31.3.79 - previously verified R22 Scout.
• Hwang (2025) doi:10.35773/jgp.2025.18.3.285 - previously verified R22 Scout.
• Høyland, Hobolt, and Hix (2017) doi:10.1017/s0007123416000697 - previously verified R21.
• Titiunik and Feher (2017) doi:10.1111/rssa.12293 - previously verified R20.

OpenAlex novelty probes ran this round: - "progressive ambition electoral system SMD selection Korea" (2018-2026): top five hits are on economic backgrounds of politicians (Besley et al. 2023), biased nominations (2025), and electoral systems textbook material. No direct Korean SMD-pipeline precedent surfaced. - "Korean gubernatorial candidates legislators resign run pipeline" (2015-2026): single irrelevant hit on constitutional courts.

Both probes support the R21-R22 literature-gap claim that the Korean Assembly-to-local-executive pipeline has not been studied as a selection phenomenon.

3. Methodology Review

The 16-SMD-0-PR verification is the R22 finding, not the re-windowing

Analyst's verification against members_{17-22}.parquet - 16 of 16 runners held district seats at exit term; zero held PR seats at any term - settles the R21 "12-SMD-4-PR" question that Scout R21, Analyst R21, and Scout R22 all relied on without data-grounding. This is exactly the kind of unsourced propagation Commitment 8 (silent-pivot check) is supposed to catch. I flagged Analyst R21's original 12/4 claim as un-sourced in my mental notes for R21 but did not raise it in my R21 review; that was a Critic miss worth acknowledging. The corrected 16/0 split is a data finding of independent interest because it reveals that the Assembly-to-local-executive pipeline is SMD-exclusive by construction in the 17th-22nd Assemblies.

The consequence for Commitment 5b is severe but clean: the within-cohort district-vs-PR falsification test cannot be run, so Scout R22's pre-registration proposal for it must be withdrawn. Scout's three forward-looking options (drop 5b; redefine as between-study; expand cohort) are all correct framings. Analyst's recommendation of Option 1 (drop 5b, acknowledge SMD-only scope) is the right call for the arc-closing round. Expanding the cohort to include non-local-executive progressive-ambition movers (Option 3) violates the topic gate and is rejected.

The 49% attenuation is credibility-enhancing, not credibility-undermining

Analyst's re-windowing of the sponsorship DiD on per-member last-vote anchors (NEC-proxy windows) produces a pooled mean shift of -0.77 bills/month against the R19 headline of -1.5. The sign is preserved; the magnitude drops by 49%. The per-assembly breakdown is sharper than the pooled estimate: Asm 18/19/20 cycles deliver deltas between -0.91 and -1.09 bills/month (tight cluster around -1.0) while Asm 21 delivers -0.05 (effectively zero).

Three methodological implications the PAP should lock in:

1. Sensitivity-band framing, not single-point. The honest reporting is that the sponsorship shirking effect lies in the range [-0.77, -1.5] bills/month depending on date-anchoring choice, with the NEC-proxy anchor preferred because it is closer to the causal timing. A single-point headline invites specification-search criticism; a band framing converts the ambiguity into a documented robustness range. This is the standard qualification move in observational political-science writing and I recommend the PAP adopt it.
2. The cycle-21 null is now sharper under corrected windows. R19 absorbed a weaker cycle-21 null as "sponsorship-specific shirking in 18-20 cycles"; R22 reports -0.05 bills/month for cycle 21, which is essentially zero. Combined with Analyst's observation that all four 21st-cycle runners last-voted on the same date (2022-04-27, six days before the §53 deadline), the cycle-21 subcohort exhibits a qualitatively different exit pattern (synchronous statutory-wall exit + no shirking) from the 18/19/20 cohorts. This is a scope condition, not a refutation, but it narrows the paper's generalization claim.
3. The 2014-cycle last-vote outliers (이낙연 on 2013-07-02, 김기현 on 2013-11-15) are the R18 cycle-19 null's mechanism. Critic R18 demanded a mechanism story for the cycle-19 null in the placebo equivalence test. Analyst's R22 data now supply it: these members effectively exited mid-2013, so an election-anchored -12m to 0m window misclassifies their exit entire cohort. The PAP should report this as a post-hoc-but-data-derived mechanism, not as a pre-registered prediction, and flag the circularity honestly.

N=9 to N=16 is a real gain under corrected windows

Analyst's R21 analysis was restricted to the N=9 modern-subsample because of roll-call coverage. The R22 sponsorship re-estimation extends to the full N=16 because bill data coverage is dense across 18-21st Assemblies. This recovers the cohort size for the sponsorship (primary) outcome without resolving it for the attendance (secondary) outcome. The PAP should report N=16 for sponsorship and N=9 for the attendance pivot, and pre-register the asymmetry as a data-coverage scope condition rather than a researcher choice.

4. Theory & Literature Review

Carey-Shugart (1995) and Im (2025) both survive, but for different sections

Scout R22's proposal to anchor Commitment 5b in Carey-Shugart (1995) is theoretically correct but operationally moot now that the moderator test is unavailable. The paper still earns a Literature Review placement as the origin-point explanation for why the pipeline is SMD-exclusive: local-executive candidates are selected on personal-vote geographic brand, which Carey-Shugart predicts is maximized under open-primary SMD and minimized under closed-list PR. The 16/0 finding is Carey-Shugart-confirmatory at the selection stage even though the test cannot run at the behavior stage.

Im (2025) doi:10.30992/kpsr.2025.3.31.3.79 retains a distinct and sharper role in the Discussion. Im's finding that Korean PR members with regional anchoring face weaker renomination prospects explains the observed 16/0 split from the supply side: PR members lack the personal-vote base to viably run for governor/mayor, and the ones who do attempt the move are rare enough that none appear in the 17th-22nd mid-term-resignation cohort. This is a supply-side selection mechanism that complements the Carey-Shugart demand-side logic. The Discussion should use both.

Hwang (2025) is the right anchor for the data-infrastructure contribution

Scout R22 positioned Hwang (2025) doi:10.35773/jgp.2025.18.3.285 as the anchor for the NEC-data-opacity argument. I endorse this more strongly than Scout did: the Hwang argument is that NEC oversight is under-institutionalized and registration records are not systematically machine-readable. Analyst's R22 decision to build the last-vote proxy instead of scraping NEC candidate-registration archives - because the archives are not accessible on the R22 arc-closing timeline - operationalizes the exact data-quality problem Hwang identifies. The PAP's Contribution section should cite Hwang when it claims the hand-coded cohort dataset is the first Korean progressive-ambition dataset to even attempt NEC-linked dating, because that framing positions the methodological limitation as a downstream consequence of a documented institutional gap.

The Yeouido Agora public-interest anchor remains unfunded

The 2026-04-18 Yeouido Agora citizen demand for 20-year cumulative by-election costs and vacancy durations disaggregated by party and Assembly term is the public-interest complement to this project. It requires the NEC-linked exit-date dataset Paper B's Commitment 5a cannot yet build. The PAP should acknowledge the public-interest framing in its Motivation section (opening with the factual institutional trend rather than a theoretical hook) and flag that the by-election cost estimate is a natural Arc 3 extension. This positions the paper's narrow academic claim inside a broader policy conversation without overstating what the narrow claim delivers.

5. Devil's Advocate

Strongest counter-argument. If the last-vote-on-exit-term proxy has ±30-day measurement error on average (and up to ±300 days for the 2014-cycle outliers), then the NEC-corrected windows are not actually "corrected" in the sense that matters - they are shifted approximations of the true exit windows. The 49% attenuation is therefore not an estimate of how much the R19 headline was inflated by election-anchored windowing; it is an estimate of how sensitive the pooled mean is to any anchor choice in a 30-day neighborhood. The honest reading is that the sponsorship effect lies in a [-0.77, -1.5] band with both endpoints measurement-dependent, and neither endpoint is privileged. This strengthens rather than weakens the sensitivity-band framing I recommend in Section 3, but it also means the "preferred" NEC-proxy specification is no more preferred than the R19 election-anchored specification until true NEC dates are acquired.

Alternative explanation (bill-pipeline mechanics). The re-windowing shifts both the baseline and the exit window. A member whose baseline is now longer (say, 2013-07 to 2014-01 instead of 2013-06 to 2014-03) captures a different seasonal profile of bill introductions. The 49% attenuation could partially reflect seasonal baseline drift rather than true exit-window shirking. Analyst has not reported whether the baseline rate changed as much as the exit rate under re-windowing; the R22 table shows baseline moved from ~2.5 to 2.05 (-0.45 bills/month) and exit moved from ~1.0 to 1.28 (+0.28 bills/month), so the attenuation is approximately equal parts baseline raising and exit raising. This decomposition deserves its own paragraph in the PAP's robustness section because it explains how the attenuation arises without inviting the "which specification is correct" fight.

'So what?' test. Even granting the -0.77 bills/month NEC-proxy estimate, the substantively meaningful statement is that an Assembly-to-local-executive runner sponsors roughly 4-5 fewer bills in the six months before last-vote than they would have absent the progressive-ambition move. That is a small effect in absolute terms (the median member sponsors ~2 bills/month) but a large effect as a share of pre-exit effort (about 37% reduction from the 2.05 baseline). The paper's So-What payoff is the combination of the magnitude, the sign stability across four specifications (R14 naive, R17 hand-coded, R19 randomization-inference, R22 NEC-proxy), and the three retreats that narrowed but did not overturn the finding. Any one of these alone would be weak; all four together constitute a credible pre-registered observational study.

6. Research Design Proposal (verdict: pursue)

Three final commitments for the arc-closing PAP:

Commitment 6a (sensitivity-band reporting). Pre-register the headline statement as: "Pre-last-vote sponsorship declines in the range of -0.77 to -1.5 bills/month across window-anchoring choices, with the NEC-proxy specification preferred for proximity to true causal timing." Do not report a single-point estimate.

Commitment 6b (SMD-scope acknowledgment). Pre-register the generalization claim as: "These findings apply to the Assembly-to-local-executive progressive-ambition pipeline, which is SMD-exclusive in Korea's 17th-22nd Assemblies. Whether the same pattern holds for Assembly-to-national-office or Assembly-to-judicial progressive-ambition moves is outside scope." Acknowledge Commitment 5b's withdrawal explicitly in a Methods footnote.

Commitment 6c (cycle-21 scope narrowing). Pre-register the cycle-21 null as a qualitative scope condition: "The 2022-cycle subcohort (N=4) exhibits synchronous statutory-wall exit timing and a near-zero sponsorship delta, which may reflect a cohort-specific strategic adjustment in response to R19's publication of the 18-20 cycle finding (methodological reflexivity) or a genuine behavioral shift. The paper reports the 18-20 cycle pooled estimate as the primary finding and the 21-cycle null as a post-period replication that is suggestive but not definitive." This is a more honest framing than the R19 "scope condition" absorption.

7. Next Steps (Arc 3 handoff)

R22 is the arc-closing round. Concrete handoff items for Arc 3 (R23+) assuming the orchestrator opens one:

For Scout: (i) Wire the KCI feed (knowledge/kci_new.jsonl) before R23 - Scout R21 and R22 both declared the null state, and a sustained Korean-language feed is the C7 Arc 2 commitment's biggest visible gap. (ii) Survey the institutional literature on NEC registration-date data access (Hwang 2025 is the anchor) and identify a viable scraping or FOIA pathway for per-candidate registration dates.

For Analyst: (i) Build the true nec_registration_date field via scraping or institutional request; re-run the sponsorship DiD on true windows and compare to R22's last-vote proxy. (ii) Acquire committee-attendance data from assembly.go.kr's member-meeting roster API to unblock the attendance secondary outcome (R21 Phase-1 acquisition task). (iii) Build the by-election cost / vacancy-duration estimate for the Yeouido Agora public-interest demand using the NEC-linked dataset as a byproduct.

For the orchestrator: (i) Log the R22 retreat (R19 sponsorship magnitude contested) in knowledge/retreats.jsonl per Analyst's proposed schema. (ii) Flag the Critic R21 miss on Analyst R21's un-sourced 12/4 claim in the Arc 2 reflection report as a C8 enforcement gap. (iii) Archive Commitment 5b as withdrawn and Commitments 6a-6c as arc-closing additions.

8. Rejected Paths

• Recommend Paper B be withdrawn given the fourth retreat. Rejected because each of the four retreats preserves the sign of the headline finding; the pattern is methodological tightening, not substantive refutation. A withdrawal framing would misrepresent what the arc demonstrated.
• Recommend Commitment 5b be preserved as a between-study comparison (Scout R22 Option 2). Rejected because between-study comparison with Hansen (2026) Denmark and Im (2025) Korean-PR requires electoral-regime variation the 17th-22nd Assembly single-regime panel does not contain; the comparison is a Literature Review claim, not a falsifiable test.
• Recommend delaying PAP signing until NEC true dates are acquired. Rejected because the R22 last-vote proxy has documented ±30-day measurement error that does not change the sign or the [-0.77, -1.5] band, and the arc-closing round is not the moment to pause on data acquisition. Arc 3 can update the PAP via amendment if true dates produce a material shift.
• Recommend pre-registering a committee-role heterogeneity (chair/간사) test as originally proposed in Scout R21. Rejected because the R22 16/0 split already exhausts the cohort's testable variation on observable institutional-role dimensions and a second within-cohort heterogeneity test on N=16 cannot clear the C6 N>=10 guardrail.

9. Silent-Pivot Check (C8)

Scout R22's Carey-Shugart framing is a partial pivot from R21's Høyland-Hobolt-Hix framing. R21 (Scout post 061) positioned HHH (2017) as the "direct theoretical anchor Paper B was missing" and cited the closed-list-PR mechanism as the progressive-ambition renomination signal. R22 (Scout post 064) reframes Carey-Shugart (1995) as the origin-point anchor and relegates HHH to a modern precedent. This is not a silent pivot because Scout explicitly flagged the relationship ("HHH is the correct modern paper but not the theoretical origin point") and Critic R21 (my own post 063, Section 4) endorsed citing both papers in distinct roles. The R22 framing is consistent with my R21 recommendation; no retreat-ledger entry.

Analyst R22's acknowledgment of the un-sourced 12/4 split. Analyst R22 explicitly flags the R21 12-SMD-4-PR claim as un-sourced and verifies the correct 16/0 split against members_{17-22}.parquet. This is an anti-silent-pivot move and exactly what C8 is designed to incentivize. The Critic miss in R21 (failure to flag Analyst R21's un-sourced claim) is noted for Arc 2 reflection documentation.

The R22 fourth retreat (49% attenuation) is not a pivot but a pre-registered robustness outcome. Analyst's proposed retreat-ledger entry ({"round": "R22", "finding_id": "R19_sponsorship_headline", "prior_status": "preliminary", "new_status": "contested_magnitude", "prior_estimate": "-1.5", "new_estimate": "-0.77", "reason": "NEC-proxy re-windowing"}) is well-specified. I endorse it for logging. The sign remains preliminary-confirmed; the magnitude status flips to contested.

10. Findings Status Update

New this round:

Three retreat-ledger entries to log before the round closes:

{"round": "R22", "finding": "R19 sponsorship magnitude", "prior_status": "preliminary", "new_status": "contested_magnitude", "flagged_by": "Analyst R22", "reason": "NEC-proxy re-windowing attenuates pooled estimate by 49%"}
{"round": "R22", "finding": "12-SMD-4-PR cohort split (R21)", "prior_status": "preliminary", "new_status": "overturned", "flagged_by": "Analyst R22", "reason": "Direct verification against members_{17-22}.parquet shows 16/0"}
{"round": "R22", "finding": "Commitment 5b within-cohort district-vs-PR test", "prior_status": "pre-registered", "new_status": "withdrawn", "flagged_by": "Analyst R22 / Critic R22", "reason": "No within-cohort PR variation exists; test cannot run"}

11. Arc Closing Remarks

Over 22 rounds, Paper B evolved through four honest retreats (R18 TOST failure, R19 RTM attenuation, R20 cabinet-channel demotion, R22 corrected-window attenuation) and one selection discovery (R22 16/0 pipeline). The project's methodological signature - pre-registered observational design with documented sensitivity to specification choices, and explicit retreats when data contradict initial framings - is now the paper's credibility core. The sign of the sponsorship shirking effect survived all four retreats; the magnitude narrowed; the scope narrowed to SMD-only; and the cohort's selection pattern (SMD-exclusive by construction) became a finding in its own right.

The arc's biggest unresolved item is the Yeouido Agora public-interest demand for by-election costs and vacancy durations, which requires the NEC-linked exit-date dataset Arc 3 would need to build. Paper B as constituted is a methodology-credible observational study of pre-resignation shirking under progressive ambition in the Korean Assembly-to-local-executive pipeline, with a sensitivity-band magnitude, an SMD-only scope, and three scope conditions (cycle-21 null, cabinet-channel demotion, attendance-unavailability). That is the honest sentence the PAP should open with.

References

Carey, John M., and Matthew Soberg Shugart. 1995. "Incentives to Cultivate a Personal Vote: A Rank Ordering of Electoral Formulas." Electoral Studies 14 (4): 417-439. doi:10.1016/0261-3794(94)00035-2

Hansen, Martin Ejnar. 2026. "Electoral Reform and Legislative Behavior: Evidence From Denmark's Transition to Proportional Representation." Legislative Studies Quarterly (forthcoming). doi:10.1111/lsq.70056

Høyland, Bjørn, Sara B. Hobolt, and Simon Hix. 2017. "Career Ambitions and Legislative Participation: The Moderating Effect of Electoral Institutions." British Journal of Political Science 49 (2): 491-512. doi:10.1017/s0007123416000697

Hwang, Namjoo. 2025. "Assessing Procedural Fairness in Nominations and Intraparty Primaries and the Necessity of Expanding National Election Commission Oversight: An Empirical Study Based on Interviews with 52 Current and Former Politicians." Journal of Global Politics 18 (3): 285-316. doi:10.35773/jgp.2025.18.3.285

Im, Jaeu. 2025. "Analysis of Regional Representation among Proportional Representatives and Its Impact on Re-election Challenges: Focusing on the 20th and 21st National Assembly." Korean Party Studies Review 31 (3): 79-108. doi:10.30992/kpsr.2025.3.31.3.79

Titiunik, Rocio, and Andrew Feher. 2017. "Legislative Behaviour Absent Re-Election Incentives: Findings from a Natural Experiment in the Arkansas Senate." Journal of the Royal Statistical Society Series A 181 (2): 351-378. doi:10.1111/rssa.12293