按Hauser和Duncan(1959)的经典划分,广义人口学(population studies)关注人口系统内部变量与外部社会经济变量之间的交互关系。其核心议题包括人口与经济(劳动力供给、代际转移、消费结构)、人口与健康(疾病负担、卫生服务需求)、人口与城镇化(空间再分布、城市体系)、人口与家庭(家庭形成、照护安排、代际关系)、人口与环境(资源约束、气候-人口互动)等。
这些议题的传统研究范式建立在两类数据基础之上:普查与抽样调查提供人口学变量的截面或短期纵向信息;行政记录(税务、社保、教育登记)提供制度运行中的人口相关信息。分析方法以回归分析、分解技术和情景预测为主,时间分辨率通常为年度或十年级。
Breen和Feehan(2025)在Population and Development Review五十周年特刊中系统梳理了正在改变这一范式的五类新型数据源:数字痕迹(digital traces)、大规模行政数据链接(administrative data linkage)、数字化历史记录(digitized historical records)、众包系谱数据(crowdsourced genealogical data)和卫星遥感数据(satellite and remote sensing data)。他们指出,这些数据源的共同特征是并非为人口学研究而设计,但可以被重新利用(repurposed)以回答人口学问题——且在某些维度上,它们使以前无法操作化的问题第一次变得可研究。
迁移(migration)研究的方法论史,本质上是一部不断扩展可观测性的历史。每一次方法论进步都回应了一种具体的"看不见"——人口空间流动中某个无法被现有工具捕捉的维度——并发展出新的技术将其纳入分析视野。
现代迁移研究始于Ravenstein(1885, 1889)对英国普查数据的分析。他从出生地与现住地的交叉表中归纳出一组经验规律("迁移法则"):大多数迁移为短距离迁移,迁移量与距离成反比、与人口规模成正比,经济动机占主导。这些观察不是理论,而是普查数据所允许的最大限度的经验概括:Ravenstein能观测到迁移的净结果(人们现在住在哪里),但无法观测过程(何时迁移、迁移几次、经由何种路径)。
前述方法论演进中,每一阶段的分析工具都依赖于"目的性采集"的数据——生命登记系统记录死亡事件,截面调查采集健康状态,纵向面板追踪状态转移。这些数据的共同特征是:它们是人口学家为了回答人口学问题而设计并实施的数据采集活动的产物。
数字化时代的根本变化不在于某种具体技术的出现,而在于数据来源的认识论转变:从"向人群提问"转向"从人群的数字痕迹中提取信息"。电子健康记录(EHR)是为临床诊疗产生的,手机信令数据是为通信计费产生的,可穿戴设备数据是为个人健康管理产生的——它们都不是为人口学研究而设计的,但都可以被重新利用(repurposed)为人口学分析的输入。
这一转变带来了三个层面的结构性后果。
The history of migration research within formal demography is, at its core, a history of expanding observability. Each major methodological advance has responded to a specific form of invisibility — an aspect of human spatial mobility that existing tools could not capture — and has developed new techniques to make that aspect amenable to systematic analysis. This narrative of progressively "seeing more" provides the organizing logic for the present review.
Modern migration research begins with Ernst Georg Ravenstein's analysis of British census data, published in two landmark papers in the Journal of the Statistical Society of London (1885) and the Journal of the Royal Statistical Society (1889). Working with birthplace data from the 1871 and 1881 censuses, Ravenstein identified a set of empirical r
生育与死亡共享人口学分析的基本结构——年龄别比率、合成队列指标、生命表逻辑——但在本体论上存在根本差异:死亡是不可逆的吸收事件,而生育具有可重复性、序列性和时机选择性。这一差异决定了生育方法论的演进轨迹远较死亡研究复杂。
总量测量阶段。 早期生育测量依赖生命登记系统,从粗出生率(CBR)到年龄别生育率(ASFR)及总和生育率(TFR),逐步排除人口年龄结构的干扰。TFR本质上是与时期生命表同构的合成队列指标(Preston et al., 2001)。在登记系统不完善的发展中国家,"亲生子女法"(Cho et al., 1986)和"反向存活法"(Spoorenberg, 2014)利用普查数据间接估计生育率,极大拓展了分析的地理覆盖面。
过程建模阶段。 Henry(1953)引入可生育力(fecundability)概念,将生育从简单计数转化为随机等待过程;其"自然生育力"概念(Henry, 1961)则提供了衡量刻意控制程度的反事实基准。Coale和Trussell(1974, 1978)据此构建参数化模型生育率表,通过控制参数m识别人口开始实施生育控制的时间节点,直接服务于普林斯顿欧洲生育转变项目(Coale & Watkins, 1986)的核心需求。值得注意的是,Princeton EFP发现生育转变的时机与语言文化边界的相关性强于经济指标(Watkins, 1991),对纯经济决定论提出了挑战。
Fertility shares with mortality the fundamental structure of demographic analysis — age-specific rates, synthetic cohort measures, life table logic — but departs from it in a critical respect. Death is a singular, irreversible, absorbing event. Fertility is renewable, sequential, and timing-dependent: a woman may have zero, one, or several children; each birth alters the conditions for the next; and the spacing and stopping of births carry distinct demographic and social meaning. This ontological difference has driven a methodological trajectory that is both parallel to and fundamentally more complex than the one traced in mortality research.
The earliest fertility measures — the crude birth rate (CBR), defined as the number of live births per thousand population in a given year — were direct products of the same vit
The life table is the foundational instrument of formal demography. Halley (1693) constructed the first empirically grounded life table from death records of the city of Breslau, inaugurating a tradition that would define the discipline for three centuries. The life table translates age-specific mortality rates into a synthetic cohort narrative: given current mortality conditions, how many years can a newborn expect to live?
The elegance of the life table lies in its minimal data requirements — only age-specific death rates are needed — but this simplicity comes at a cost. The standard period life table rests on a stationary population assumption: it treats the cross-sectional mortality schedule of a single period as if it were the lifelong experience of a hypothetical cohort, implicitly assuming