# Amendment 01 to Pre-Registration: Divided Government and Significant Lawmaking

**Date:** 2026-07-15
**Amends:** PRE_REGISTRATION.md, frozen 2026-07-14
**Blindness attestation:** made after the stop-1 spine and gates, before any model was fit. No regression, no coefficient, no trend, and no DW-NOMINATE merge existed when this decision was taken. The decision rests on the construction of the measure, not on any result it produces.

---

## 1. What the frozen spec said

Section 3 named the primary DV as Mayhew's combined ("Sweep Two") significant-legislation total per Congress.

## 2. What stop 1 found

The list is not one homogeneous series. In Mayhew's own words (2013-14 file, p.2), for Congresses since 1990 he fell back entirely on the Sweep One procedure, bypassing Sweep Two, because the retrospective sweep cannot be performed for recent years.

- 1947-1990 (80th-101st): two-sweep combined. Sums to 267, matching Mayhew's published total, which authenticates the file.
- 1991-2024 (102nd-118th): Sweep One only.
- Of the 267 combined pre-1990 laws, 211 carry a Sweep-One flag and 56 (21%) are Sweep-Two-only.

The frozen primary DV is therefore not merely inconvenient after 1990, it is undefined. No combined total exists to extract.

## 3. The amendment

**Primary DV is now the Sweep-One-only count per Congress, constructed consistently across the full 80th-118th span.**

The mixed series (combined pre-1990, Sweep One after) with a pre/post-1990 regime indicator is retained and reported as a robustness DV.

## 4. Why, stated before any estimate

The seam is confounded with the key predictor. Sweep Two inflates pre-1990 counts by roughly 21% and exists only before 1990. Divided government became markedly more common after 1994. A measurement procedure that mechanically lowers counts after 1990 therefore coincides with divided government becoming more common, so on the mixed series the divided coefficient would partly absorb "counted by a different procedure."

This is the same failure the mandatory time trend in section 5 was registered to prevent, in a more dangerous form: the seam is a step at 1990, not a smooth trend, so a linear term does not absorb it, and a regime indicator for it is close to collinear with the era the study is about.

Mayhew's own fallback to Sweep One is the operative precedent. Sweep One is the only procedure applied across the entire span, so it is the only series in which 1955 and 2015 are counted the same way.

## 5. Cost of the amendment, stated plainly

- The Sweep-One-only series discards 56 pre-1990 laws that Mayhew judged significant on retrospective review. The primary series is thus a contemporaneous-judgment measure throughout, not a claim that those 56 laws were unimportant.
- Sweep One before and after 1990 is the same procedure, but not guaranteed to be identically applied across decades. Residual measurement drift within Sweep One is possible and is not corrected.

## 6. Reporting rule (binding)

Both series are reported. If the divided coefficient agrees in sign and rough magnitude across the Sweep-One-only primary and the mixed-with-regime-indicator robustness, the finding is reported as robust to the seam. If they disagree, the disagreement is the finding and is reported as such. Neither is dropped, and the friendlier one is not promoted.

## 7. Other corrections from stop 1 (recorded, not amendments)

- **Span.** The verified endpoint is the 118th Congress (2023-24), from Mayhew's file dated 2025-01-06. The frozen spec anticipated 2014, 2018, or 2022 and set the rule "use the last fully covered Congress," which this satisfies. Span is 80th-118th, n = 39, not the "roughly 38" the spec estimated. The endpoint rule is unchanged; only the stale external estimate was wrong.
- **Prior literature.** Ansolabehere, Palmer, and Schneer report roughly four fewer significant acts under divided government in the 20th century, so the prior literature is not uniformly null. Section 1's registered expectation (null) is unchanged, but the both-ways test across measures now carries more weight and APS disagreement is a substantive result, not noise.
- **107th Congress.** Mayhew and the executive-orders control table agree it is divided. No maker-checker conflict arose.