Seven Seats, Five Maps 五张图，七个席位

Congressional districts in the United States are redrawn once every ten years, after the decennial census. That cadence is the rule the country has run on for most of its history. In the past nine months five states — Texas, California, Missouri, North Carolina, and Virginia — broke it. The latest count is easier to read as two opposing columns: Texas, Missouri, and North Carolina aim at roughly seven additional Republican seats; California and Virginia aim to offset roughly six of them.

Read the figure left to right. Texas's legislature redrew thirty-eight districts to convert roughly five seats from blue to red. California passed Proposition 50 to suspend the state's independent redistricting commission for one cycle and redrew fifty-two districts to reverse Texas's gain. North Carolina, already carrying one of the most gerrymandered maps in the country since 2023, tightened it further. The strip above the maps now makes the accounting explicit: this is not one neutral update cycle, but a set of mid-cycle interventions large enough to swamp the current three-seat House margin.

The temptation is to score these maps by how strange they look — Polsby–Popper compactness scores, ratio of perimeter to area, the old "if a district resembles a salamander, it's a gerrymander" heuristic. Mathematicians have spent the last decade demonstrating that this heuristic doesn't work. A district can hug a coastline and look ugly while being perfectly fair; another can be a tidy rectangle and be drawn to crack a city in half. The better test, brought before the Supreme Court in Rucho v. Common Cause (2019), is statistical: generate tens of thousands of legal maps by Markov chain Monte Carlo, plot the partisan outcomes, and see whether the proposed map sits at the extreme tail of that distribution. Several of the maps shown here do — at the 99th percentile or beyond. Yet Rucho closed federal courts to partisan-gerrymandering claims. The statistical evidence can be sharp while the federal remedy remains blocked.

What remains is structural. Single-member, winner-take-all districts are the soil that gerrymandering grows in: a few thousand voters moved across a line can move a seat. Cambridge, Massachusetts; Eastpointe, Michigan; Portland, Oregon — three places that have switched their local elections to ranked-choice voting in multi-member districts — show what a country less hospitable to this kind of map would look like. Until that change reaches the House, the figure above is what the country looks like when both parties decide that the old once-a-decade rule is no longer binding.

A note on the boundaries. Texas's "new" panel uses the PlanC2308 boundary file; California and North Carolina still use 118th Congress district geometry as a visual base because comparable new boundary files were not available in this build. Their colour and R/D counts show the 2026 projection, not a claim that every polygon line is final. Seat counts are best public estimates as of the article's date and may move as litigation and implementation settle.

按惯例，美国国会选区十年一划，等人口普查结束再调整。大多数时候， 这套节拍还算稳定。可过去九个月里，德州、加州、密苏里、北卡和 弗吉尼亚先后半途重划。现在更清楚的读法是两列账：德州、密苏里、 北卡合计预计把约 7 个席位推向共和党；加州和弗吉尼亚则试图替 民主党抵回约 6 个。

读图可以从左到右。德州州议会重划 38 个选区，预计把约 5 个 民主党席位推向共和党；加州通过 Proposition 50，暂停独立划区 委员会一个周期，重划 52 个选区，试图把德州多出来的红色优势抵回去； 北卡的图自 2023 年以来已经高度偏向共和党，这次又进一步收紧。 上方席位条把这笔账摊开：这不是普通的周期性调整，而是一组足以 淹没当前 3 席多数优势的半途干预。

看这类图，很容易先问它丑不丑：紧凑度、周长面积比，或者那句 老话——选区像蝾螈，就是 gerrymander。但过去十年的数学研究反复说明， 丑不是重点。沿海选区可能形状很怪，却并不偏；一个整齐矩形也可能 刚好把城市选民拆散。更有力的检验是统计：用马尔可夫链蒙特卡罗 生成成千上万张合法地图，看党派结果会落在哪个分布里，再问眼前 这张是不是跑到极端尾端。2019 年的 Rucho v. Common Cause 案把这种方法带到最高法院；这里几张新图也正落在 99 百分位甚至更外。 问题是，Rucho 同时让联邦法院退出了党派操弄划区诉讼。 证据可以很清楚，路却被堵上了。

更深的麻烦在制度本身。单席位、赢者通吃的选区最适合被切割： 几千张票跨过一条线，就可能换掉一个席位。马萨诸塞州剑桥、密歇根州 Eastpointe、俄勒冈州波特兰已经在地方选举中采用多席位选区和 排序选择投票；它们说明，代表制度可以设计得不那么容易被几条线 操弄。在这种改革进入众议院以前，只要两党都觉得 十年一划不必再守，地图就会长成上面这样。

关于边界。 德州“新图”使用 PlanC2308 边界文件；北卡和 加州仍以第 118 届国会选区作为视觉底图，因为这一版构建还没有 纳入可比的新边界文件。它们的颜色和 R/D 计数表达的是 2026 预测， 不是说每一条多边形边界都已经定稿。席位估计按发稿日可见公开资料整理， 后续诉讼和执行过程仍可能改变这些数字。