The Committee-Chair Confound is Refuted (In the Opposite Direction), the Slope-Based Placebo Clears, and the Gender Gap Holds Under the All-Members Baseline: R16 Residual Tests

Critic's R15 (045_critic.md) raised three residual identification problems and Scout's R16 (046_literature_scout.md) anchored them to a specific Korean literature. This round I execute three of Critic's five priorities: the slope-based selection placebo (Priority 3), the committee-chair tabulation (Priority 2), and the all-sitting-members gender baseline (Priority 4). I also deliver a first pass on Scout's primary-timing suggestion and surface a descriptive statistic relevant to the Yeouido Agora by-election-cost demand. Two results are decisive; one is underpowered but directionally consistent.

Cohort reconstruction

I re-built the treated cohort on a cleaner definition than R14 used: a legislator is treated in a given cycle if their last chief-sponsored bill in the 24-month pre-local-election window falls between 180 and 30 days before the local-election date AND they have zero chief-sponsored bills in the post-LE remainder of that Assembly (a hard no-continuation filter). This drops cabinet-appointment slips that Critic flagged in R15 because those legislators typically keep sponsoring briefly after appointment. The replacement cohort is N=35 across 18th-21st cycles (6/14/8/7), with a continuing-colleagues pool of N=1,093. Code and cohort files at /tmp/r14_cohort.py, /tmp/treated_r14.csv, /tmp/pool_r14.csv.

Test 1: Slope-based placebo in the [-24, -12] window (Critic Priority 3)

Critic's subtler selection alternative says that legislators on an upward sponsorship trajectory (new chairships, primary wins) are both more recruitable and more likely to show a rising bill count. A slope-based check compares the change in monthly sponsorship rate from [-24, -18] to [-18, -12] for treated and pool.

Difference in slopes: -0.069 bills/month, Welch t=-0.293, p=0.771. Treated and continuers are on statistically indistinguishable pre-pre trajectories. The subtler selection-on-slope alternative does not survive. By-cycle breakdown does confirm the 18th Assembly outlier one more time: treated there are on an UPWARD slope (+0.528) while pool declines (-0.214), a within-cycle differential of +0.742. The three "anti-shirking" cycles (19th-21st) show treated slopes that are flatter or marginally steeper down than pool - not the steep upward pattern a chairship-momentum story predicts.

Test 2: Committee-chair confound runs the wrong direction (Critic Priority 2)

Critic's R15 Devil's Advocate argued that committee chairs are over-represented among resigner-candidates and mechanically over-represented among chief sponsors, confounding the position-taking story. I tried two proxies. The ppsr_kind = 위원장 (committee-chair-proposed bill) field does not identify individual chairs because chair-submitted bills record the committee as institutional proposer with rst_mona_cd null (0/1,506 matched). The cleaner proxy, following Jung (2018) and Jeon (2022)'s Korean committee-chair-allocation findings, is seniority: Korean standing-committee chairs are overwhelmingly 3선+ (three-term or more).

Fisher exact: odds ratio 0.286, p=0.0016. Treated legislators are roughly half as likely to be senior as continuers. This runs in the OPPOSITE direction from what the committee-chair confound requires. If anything, the resigner-cohort is a junior-heavy sub-population, consistent with the Mayhew (1974) credit-claiming story (junior members have more to gain from visible position-taking on the way to a bigger office) and inconsistent with chairship-mechanics. This is the decisive finding of the round. Scout's R16 prediction (using Jung 2018 and Jeon 2022's Korean baselines rather than US-Congress baselines) is vindicated.

Test 3: Gender crowd-out under the all-sitting-members baseline (Critic Priority 4)

Critic's R15 objected that the R15 gender test used productive sponsors (>=2 chief bills) as the baseline, and that the cleaner comparison is all sitting NA members. I re-ran the test against both baselines.

The two baselines are essentially identical (16.70% vs 16.93%) so Scout's R16 prediction that the finding would become starker under the all-members baseline was overstated empirically - Korean women NA members sponsor at roughly the same rate as men once in office, so the productive-sponsor filter does not compound the gender selection Kim (2004) documented. But the directional result holds: treated cohort is about half the baseline female share. The failure to reach significance is a power problem (N=35), not a pattern problem. A full test requires the NEC general-candidate pool as the comparator, which needs the registry linkage Critic has flagged as blocking in R14 and R15.

Test 4: Scout's primary-timing proxy split (21st Assembly only)

Scout's R16 suggestion, citing Kim (2025), was to split the 21st-cycle treated cohort by primary-conclusion timing. Without confirmed primary dates, I used last-bill date as a proxy: before-March-1 last bill suggests early campaign commitment (primary nominee committed in January-February), on-or-after-March-1 suggests late commitment. Of 7 treated in the 21st cycle, 4 had last bill before March 1, 2022 and 3 had last bill in March-April-May 2022. N is too small to run Test 1 separately on each subgroup; this is a placeholder awaiting the NEC primary-registration data.

A descriptive by-product relevant to Agora demand 1

The 6·3 local-election Agora brief asked for 20-year cumulative by-election costs from mid-term NA resignations. My R16 cohort provides a lower-bound count of by-election-triggering resignations: 32 SMD treated members across four cycles (6/13/8/5), with 3 PR members who would have triggered list-replacement rather than by-elections. This is a floor, not a ceiling, because my treatment is defined on last-bill timing and misses quiet resigners who tapered off earlier. The 19th-cycle count (13) is an outlier that merits separate attention; scraping 중앙선거관리위원회 선거비용 공시 for these 32 districts is a self-contained descriptive exercise that would answer the Agora citizens directly.

Party breakdown of treated (pooled): 새누리당 11, 더불어민주당 5, 한나라당 3, 민주통합당 3, 미래통합당 3, others 10. The pipeline is roughly balanced between conservative and progressive parties once Lee Myung-bak-era and Park Geun-hye-era cohorts are combined, which lets the final paper report by-election costs without appearing partisan.

District type breakdown also speaks to Agora demand 2 (metro vs non-metro vacancy duration): treated cohort is 40.0% metro vs pool 56.1% metro. Non-metro districts are over-represented among the resigner-cohort, consistent with the Korean pattern that ambitious legislators from smaller constituencies trade up to provincial governorships.

Updated status of R15 concerns

Data gaps still blocking

1. NEC resignation date and candidate registry (Critic R14 Priority 1, R15 Priority A): still blocking. The no-continuation filter partially substitutes for formal resignation dates and removes the cabinet-appointment slippage, but the treatment date remains endogenous.
2. Speech intensity event study (Critic R15 Priority C, still highest-value next outcome): hearing_meetings_summary.parquet has meeting-level aggregates but not member-level speech counts. The kr-hearings-data full speeches.parquet is a 1.1GB separate download and would enable the non-mechanically-anchored outcome Critic requested.
3. 17th-19th cosponsorship coverage (R15 Test 1 extension): unchanged from R15.
4. Committee-chair direct flag: the ppsr_kind = 위원장 proxy fails because chair-proposed bills record institutional, not individual, proposer. The seniority proxy works but is indirect.

What Critic should evaluate for theoretical framing

The seniority finding is strong enough to carry theoretical weight on its own, and it reframes the position-taking story in a useful way. Resigner-candidates are disproportionately junior, which inverts the Volden and Wiseman (2024) legislative-effectiveness prediction that the most effective legislators (typically senior) should be the most recruitable for higher office. In Korea, the pipeline selects for AMBITION rather than EFFECTIVENESS - junior members with more to gain from visibility. This is a scope condition for US-imported effectiveness-benchmark arguments, and it sharpens the Mayhew (1974) / Hansen and Treul (2015) credit-claiming framing considerably.

Two remaining theoretical loose ends worth flagging. First, does the junior-heavy pattern travel to cabinet appointments? Critic's R14 concern about 추경호/조태용 cabinet contamination suggests cabinet-appointment resigners are SENIOR not junior, which would be a clean separation of the two exit pathways on a theoretically meaningful dimension. Second, the non-metro over-representation (40% vs 56%) raises the question of whether provincial legislators face a weaker intra-party electorate for the primary and therefore face different credit-claiming incentives than metro members - a scope-condition for the position-taking mechanism that can be tested within the existing 32-SMD sample.

References

Hansen, Michael E., and Sarah A. Treul. 2015. "Aiming Higher: The Consequences of Progressive Ambition among MPs in European Parliaments." European Political Science Review 7 (3): 373-395. doi:10.1017/s1755773915000260

Jeon, Jin-young. 2022. "The Keys to Legislative Success in the National Assembly of Korea: The Role of the Majority and Minority Parties." Korean Party Studies Review 21 (4): 75-102. doi:10.30992/kpsr.2022.12.21.4.75

Jung, Jin-wung. 2018. "A Study on the Pattern of Leadership Formation in Korean National Assembly: Focusing on the Allocation of the Speaker and Chairmanships of Standing Committees." Journal of Parliamentary Research. doi:10.18808/jopr.2018.2.2.

Kim, Tae-hyoung. 2025. "Why did the two major Korean parties converge in their candidate selection methods? Focusing on the 2014-2022 local elections." Korean Journal of Political Science 33 (3). doi:10.34221/kjps.2025.33.3.8

Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven: Yale University Press.

Thomsen, Danielle M., and Aaron S. King. 2020. "Women's Decisions to Run for Office: Breadwinning, Motherhood, and Candidate Emergence." American Political Science Review 114 (4): 989-1000. doi:10.1017/s0003055420000970

Volden, Craig, and Alan E. Wiseman. 2024. "Legislative Effectiveness in the American States." American Political Science Review. doi:10.1017/s0003055424000042