你大概有过这种时刻：在住了几年的城市里拐进一条没走过的街，忽然觉得自己不该在这儿。没人说什么，也没发生什么，就是店铺的密度、走路的速度、谁在朝谁看，全换了一套。
城市把人按地址分好组，然后这一组一组的人过着差别巨大的日子。这一页拆三件事：最早那批研究者怎么看、又看漏了什么；「住在哪」到底有多大力气；以及一个街区涨价时，原来的人究竟去哪了——最后这问的答案跟直觉不一样，差别出在口径上，不在立场上。
1920 年代的芝加哥，人口几十年翻了好几番，新移民挤进城中心那圈旧房子。芝加哥大学有位 Robert Park，进学界前当了十几年报社记者，对学生的要求出了名：别坐在图书馆里，去街上，把裤子坐脏。学生真去了——跳舞厅、廉租旅馆、帮派的地盘。这批研究后来叫芝加哥学派，直觉很好用：城市不是一堆房子，是一台把人分配到位置上去的机器。1925 年 Ernest Burgess 画的同心圆图里，紧挨商业区的一圈叫过渡带——最廉价、最拥挤、流动性最高。
更早一点，德国社会学家 Georg Simmel 在 1903 年说，都市人会长出一种厌倦态度（blasé attitude）：刺激太多，神经没法样样都反应，索性把差别调平——不是冷漠，是带宽用完了。1938 年 Louis Wirth 把这一路收成三个变量：规模、密度、异质性——人越多、越挤、越不一样，熟人间的关系就越让位给按功能打交道的关系。
问题出在下一句：他们把过渡带那些拥挤的穷街区判成了「社会解组」——没有规范、没有组织、只剩散沙。
把这句话按在地上摩擦的是个二十多岁的年轻人。1937 到 1940 年，William Foote Whyte 搬进波士顿北区一个意大利裔社区住了三年半（书里化名 Cornerville），跟着一个叫 Doc 的人和他那伙街角青年混。他发现这里的秩序细得可怕：谁该请客、谁先开口、谁欠谁人情，都有稳定的位次。
最漂亮的一笔是保龄球。Whyte 把分数一场场记下来，发现球打得好不好跟帮里的地位对得上——地位低但球技好的人，在高位者面前就是打不出来，旁边人起哄、他自己也泄气。同一双手，换个场合分数就变。地位不是抽象说法，它当场把球打歪了。
可迁移的那条是：所谓「没有秩序」，很多时候是「用错了看秩序的方法」——观察者带着自己那套秩序的长相进去，找不到就说没有。（1962 年 Herbert Gans 在同城另一个区得出同样的结论，书名直接叫《城市村民》。）「社会解组」这个词后来挨了大量批评，也确实该挨：它把一群人的生活方式判成了病理。证据强度：同心圆图只是对一座城市的描述性归纳；Whyte 和 Gans 是单点长期田野，给机制不给分布。
EN In 1920s Chicago, where the population had multiplied within decades, Robert Park — a newspaper reporter for over a decade before he turned academic — sent his students out of the library and into dance halls, flophouses and gang turf. The resulting Chicago School treated the city as a machine that sorts people into positions; Ernest Burgess drew the concentric-zone diagram in 1925, with a cheap, crowded, high-turnover "zone of transition" ringing the business district. Earlier, Georg Simmel (1903) had described the blasé attitude of the metropolitan mind: too much stimulus, so the nervous system flattens differences — not coldness, exhausted bandwidth. Louis Wirth (1938) boiled the tradition down to three variables: size, density, heterogeneity. The trouble was their next step, declaring the crowded poor districts "socially disorganised". William Foote Whyte demolished that by moving into an Italian-American neighbourhood in Boston's North End from 1937 to 1940 — three and a half years alongside a man he called Doc. The order there was dense, not absent. His finest detail is bowling: he recorded scores game by game and found they tracked rank within the group, so a low-status man who bowled well elsewhere simply could not perform against his seniors. Herbert Gans found the same in 1962 and called his book The Urban Villagers. The transferable point: "no order here" is usually "wrong instrument for seeing order". The disorganisation vocabulary deserved the criticism it got, since it read a way of life as pathology. Evidence: the concentric-zone model is descriptive induction from one city; Whyte and Gans are single-site long-term fieldwork giving mechanism, not distribution.
同一座城市里，不同邮编的人预期寿命能差十几年。差距谁都测得出来，难的是下一步：是街区造成了这些人的境遇，还是境遇本就不同的人被分到了不同街区？这不是抬杠：住在哪从来不是随机的——收入、工作、亲戚在哪、能不能拿到贷款，全都参与了这个分配。所以「穷街区的孩子后来也穷」什么也没证明。
第一条路是把街区本身的性质测出来。1995 年，Robert Sampson 和 Stephen Raudenbush、Felton Earls 在芝加哥访了 8782 名居民，分属 343 个邻里群，量的东西叫集体效能（collective efficacy）：邻居之间信不信得过，加上愿不愿意为公共的事出手——看见几个半大孩子上学时间在街上晃，有没有人会去说一句。
结果发表在 1997 年的《科学》上：集体效能高的邻里暴力水平明显更低；把贫困程度、居住稳定性、个人特征和此前的暴力水平都控制掉，关联仍在。但它是相关性研究——效能低也可能是暴力的果。
第二条路硬得多：直接随机。1994 到 1998 年，美国住房与城市发展部做了个叫「搬向机会」（Moving to Opportunity）的实验，在五座城市招募 4604 户住在高贫困公租房社区的家庭随机分组，其中一组拿到只能用于低贫困社区的租房券。社会科学里很罕见——「住在哪」真的被随机化了。
早期结果让很多人失望：2007 年 Kling、Liebman 和 Katz 报告，大人的就业和收入没有改善。转折在 2016 年——Chetty、Hendren 和 Katz 把实验名单接上税务记录，等孩子长到二十多岁再看：搬家时不满 13 岁的，二十五六岁时年收入比对照组高约 3477 美元（对照组均值 11270 美元，即高出约 31%），上大学比例也更高；搬家时已过 13 岁的，收益是零甚至略负。
这一下把问题的形状换了。街区不是「有影响／没影响」的开关，更像按年累积的剂量：住得越久攒得越多，十几岁才换就攒不够了。早期看不到效应，是看得太早、也看错了人。
诚实交代两条。样本是自愿申请搬迁的高贫困公租房家庭，外推要小心；更要紧的是——我们知道有效，但不太知道靠什么有效：学校？同伴？治安？至今没有定论，而不同答案指向完全不同的做法。
EN Life expectancy can differ by more than a decade between postcodes in one city. The hard part is not measuring the gap but deciding whether the neighbourhood caused it or merely received people whose trajectories already differed — nobody is randomly assigned an address. One route is to measure the neighbourhood itself. In 1995 Robert Sampson, Stephen Raudenbush and Felton Earls surveyed 8,782 residents across 343 Chicago neighbourhood clusters, measuring collective efficacy: trust among neighbours plus willingness to intervene for the common good. Their 1997 Science paper found violence markedly lower where collective efficacy was high, holding poverty, residential stability, individual characteristics and prior violence constant — influential, but correlational, and the causal arrow could run backwards. The harder route is randomisation. From 1994 to 1998 the US Department of Housing and Urban Development ran Moving to Opportunity, enrolling 4,604 families from high-poverty public housing in five cities, one randomised group receiving a voucher usable only in low-poverty areas plus counselling. Early results disappointed: Kling, Liebman and Katz reported in 2007 that adult employment and earnings did not improve. The turn came in 2016, when Chetty, Hendren and Katz linked the roster to tax records: children who moved before age 13 earned about $3,477 more per year in their mid-twenties against a control mean of $11,270 — roughly 31% — and attended college at higher rates, while those who moved after 13 gained nothing. A neighbourhood is not a switch but a dose accumulating by the year. Two honest caveats: the sample is families who volunteered to move, and we know it works without knowing through what — school, peers, safety remain unsettled, and different answers imply very different policies.
1964 年，在伦敦做研究的社会学家 Ruth Glass 给她看到的事造了个词：工人阶级住的一排排旧房子被中产一栋栋买下、翻新，整片的社会构成随之换掉。她管这叫士绅化（gentrification）。它由什么驱动，一直有两套解释并存。
一套看资本：1979 年 Neil Smith 提出租隙（rent gap）——老城区当前收到的地租远低于它本可以收到的，差额够大，资本就流进来翻新套利，谁搬进来是结果不是原因。另一套看人：David Ley 说是需求侧，战后长出一批做文化与知识工作的新中产，要老房子、要步行可达，是他们的偏好把资本引过去的。两边证据都拿得出手，谁也没压倒谁——争的是权重，不是有无，本页不替任何一边下判断。
真正反直觉的在下一层。直觉上士绅化＝原住民被赶走。但从 2000 年代起，一批追踪具体个人的研究给出了别的结果：Freeman 用全国住房调查数据、Ding 等人用费城的信用局面板、Brummet 和 Reed 用人口普查纵向数据——士绅化街区里低收入居民的迁出率，普遍不比相似的非士绅化街区更高。
先别下结论：同一批研究还发现，街区人口构成之所以变，主要来自迁入端——受过大学教育的人搬进来的速率高得多。
这两件事放在一起，说明迁出率只对准了一半画面。房子腾出来是常事——换工作、结婚、回老家。真正决定街区变不变的是它被谁接手：接手的都是付得起新价钱的人，原住民的下一代就再也进不来了。Peter Marcuse 在 1980 年代给这种情况起过名字：排斥性位移（exclusionary displacement）——被挡住的人从不出现在任何迁出名单上，因为他们根本没来过。
所以这里的分歧首先不是立场之争，是口径之争：一个测流出，一个要测本来会有的流入。同一批研究还发现：迁出的低收入居民搬去的地方往往更差，租户里教育程度较低的受冲击更大。「总体迁出率不高」和「某些人被打得很重」可以同时成立。
证据强度：这几项是行政数据或信贷面板，能跟踪同一批人，比横截面强，但都不是随机化实验，不同城市差别很大。至于该不该干预，那是价值判断，不在这门学科的量程里——这里能说清的只是：你要改的是哪一根箭头。
EN In 1964 the sociologist Ruth Glass coined "gentrification" for what she saw in London: working-class terraces bought up, refurbished, and the social composition of whole districts replaced. Two explanations have coexisted ever since. Neil Smith's rent-gap theory (1979) looks at capital: when the rent a decayed district currently earns falls far enough below what it could earn, investment flows in to capture the difference, and who moves in is consequence rather than cause. David Ley's account looks at demand: a post-war professional and cultural middle class whose tastes ran to old housing and walkable districts drew the capital after them. Both have solid evidence; the dispute is over weights, not existence, and this page takes no side. The counterintuitive layer is below. Studies that track individuals — Freeman on national housing survey data, Ding, Hwang and Divringi on a Philadelphia credit-bureau panel, Brummet and Reed on longitudinal census data — generally find that low-income residents of gentrifying neighbourhoods do not move out at higher rates than comparable residents elsewhere. The same studies find that compositional change comes mostly from the inflow side: college-educated in-migrants arrive at much higher rates. So an out-migration rate captures half the picture. Homes empty routinely; what decides whether the district changes is who takes them next. Peter Marcuse named this in the 1980s: exclusionary displacement, where the excluded never appear on any list because they never arrived. The disagreement here is first of all about measurement aperture, not values. The same work also finds that those who do leave land in worse places, and that less-educated renters are hit hardest: a modest overall out-migration rate and severe harm to specific groups are both true at once.