跳至正文

教育真的测得准吗

目录

机制裁决第 123 篇 · 对称双向第 118 篇 · section E 元尺（与 CVSS/计量学同族）· 全库第 181 篇 本篇审的是「教育测量」这个名号背后承重的三件事：排名、归因与刻度。先读三句红线：

  1. 本篇不裁决任何国家、学校或教师个体的水平高下，不给「该不该排名」「该不该打分」的建议，也不对任何单一考试结果的真实含义做判断——只审「教育质量/教师优劣」这句话从分数里读出来时承重承不承得住。
  2. 「PISA 分数」与「教育质量」是两件不同的事：前者是关于一批 15 岁学生某天某场测试作答的抽样估计，后者是教育系统所有维度的总称。本篇对称审计「排名就是教育质量」与「教育测量全是数字游戏」两个方向。
  3. 全篇承重句均给出可点击来源；英文逐字引用一律标注「一手逐字」（全文取回）或「摘要逐字」。本篇沿用的官方换算（PISA 分数与「一年学习」的关系）全部来自 OECD 官方文件，算法与方法写在第五、六节，可复现。

零、一句话裁决

教育测量是真实运转、被全世界采用的制度（PISA 2022 覆盖 81 个国家与经济体、约 69 万名学生、代表约 2,900 万 15 岁学生；NAEP 是 1969 年以来美国唯一全国性持续评估；增值模型被美国数十州写入教师评估法规）——真的不是「PISA 排名约等于教育质量排名」（OECD 官方逐字声明「无法确定所有国家与经济的精确排名」、排名区间可以很宽、官方还专门写了一段「不应被考虑」的差异）和「教育测量全是数字游戏」（PISA 有测量学根基、NAEP 有纵向趋势、增值模型有 Chetty 等的正面因果证据）这两层被声称的定论。 把一轮抽样测试的分数差读成国家或教师的优劣，是把「一次作答」读成「整个系统」，把「估计值」读成「精确值」，把「相关」读成「贡献」。

一、本篇测什么：三跳升格与七件可判定的事

本篇审的不是「考试好不好」，而是「『教育质量』这个名号从被测量到被使用，中间经历了哪些跳」。拆成七件可判定的事：

  1. 尺子真不真：PISA 是什么、样本怎么抽、分数怎么算、官方怎么定义它测的「素养」（OECD PISA 2022 官方框架、技术报告、FAQ）——这是守真锚。
  2. 排名真不真：PISA 排名有没有精确可言、位次差几分算真差异、官方自己怎么看待排名（OECD PISA 2022 官方结果第一卷「排名」段逐字、NCES 官方比较口径）——这是刻度跳。
  3. 换算真不真：PISA 分数差到底能不能换算成「一年学习」「失去多少学习」（OECD 官方《15 岁学生一年学多少》论文、PISA 2018 官方报告「分数是什么意思」段）——这是刻度跳的核心。
  4. 归因真不真：增值模型（VAM）能不能把「学生分数变化」归因给「教师」、误差有多大、年度稳定性如何（Chetty-Friedman-Rockoff 2014 全文、ASA 2014 声明、McCaffrey 2009、Rothstein 2009/2010）——这是归属跳。
  5. 标准真不真：NAEP 的「Proficient」到底是什么意思、是不是「年级水平」（NAGB 官方澄清、NCES 官方 FAQ）——这是名号跳。
  6. 消费真不真：PISA/NAEP/VAM 在真实制度里怎么被用（各国教育部 PISA 冲击、美国州级 VAM 写入解雇/奖金/执照法规、teaching to the test 与作弊）——这是消费账。
  7. 反向真不真：教育测量是不是有真价值的制度（PISA 推动政策、NAEP 纵向趋势、Chetty 正面因果、效度传统）——这是反虚无账。

结构胎记：名号跳 × 归属跳 × 刻度跳，附反向红跳。

  • 名号跳：把「PISA 数学素养分」这个关于「一批 15 岁学生某次测试作答」的抽样估计，读成「国家教育质量」甚至「国家未来竞争力」——PISA 自己的定义是「素养」而非「教育质量」，NAEP 官方把「Proficient」反复澄清为「不是年级水平」。
  • 归属跳：把「学生这一年分数的变化」读成「这位老师的贡献」——增值模型把班级组成、家庭背景、随机波动全部归给老师，ASA 官方声明「相关不是因果」、教师只占考试成绩变异 1%–14%。
  • 刻度跳：把「±20 分的抽样误差」读成「精确到一位的国家排名」——PISA 官方逐字「常无法确定所有国家与经济的精确排名」，排名区间可以很宽；官方还把数学下降 15 分换算成「四分之三学年的学习」。
  • 反向红跳：排名有误差≠教育测量是骗局；VAM 有噪声≠教师质量不可测；分数被误读≠考试无用（见第八节反虚无账）。

共用动作：把一句关于「一批学生某天某场测试的作答」的话，读成一句关于「这个国家/这所学校/这位老师好不好」的话。

与库内相关篇的分界：

  • 与 AI 评测榜篇（2026-08-01，第 174 篇）分界写死：那篇把教育测量学当成熟参照系借用（Messick 效度、ETS 考试安全、Popham teaching-to-the-test 区分教原题与教内容），审的是「AI 榜分=能力」的升格链；本篇把教育测量本身当对象——那篇是「机器榜」，本篇是「人榜」。那篇引用的 Messick/Popham/ETS 在本篇不重做，只做交叉引用。
  • 与 IQ 遗传率篇（2026-07-25，第 147 篇）分界：那篇是心理测量·个体认知差异的遗传率误读（方差比冒充命定度）；本篇是教育测量·群体比较与归因（排名与增值）。IQ 篇审「一把尺的刻度」，本篇审「两把尺被读成什么」。
  • 与计量学篇（2026-08-02，第 175 篇）分界：那篇是「尺子被做对」的正面范本（SI 定义常数、GUM 不确定度、VIM 溯源）；本篇是「尺子被当客观数用」——PISA 有自己的不确定度（标准误差、置信区间、链接误差），但媒体与政策把它当精确数。
  • 与 CVSS 篇（2026-08-03，第 179 篇）分界：CVSS 篇被归 section E 元尺（尺子被当客观数用·软件漏洞）；本篇同族（尺子被当客观数用·教育）。两篇一软件一教育，共享「被读成判断的那个数」母题，但对象与证据链完全不同。
  • 与碳核算篇（2026-08-02，第 176 篇）分界：那篇是「判据在场而口径缺席」的尺；本篇是「判据在场但误差被省略」的尺——PISA 有完整的不确定度体系，计量学传统（效度论证）也在场，缺的是使用者。

编号落位：机制裁决第 123 篇·对称双向第 118 篇·section E 元尺·全库第 181 篇（2026-08-03 全库扫描确认未被占用）。去重实测（python 全库 180 篇 1,144 万字符）：PISA／教育测量（非方法论借用）／教育评估／增值模型／value-added／NAEP／TIMSS／PIRLS／教师效能／学校问责／高利害／standardized test／教育评价 全库零命中；教育测量 4 处全在 AI 评测篇（作为方法论参照系借用，非以教育测量为对象）——邻近但不重叠；ETS 命中 86 处但绝大多数是「European Trade System」等无关缩写，仅 AI 评测篇 9.3 节是教育考试中心（那篇已做，本篇不重做）。

二、守真锚：PISA 是真实运转的全球抽样评估，先清点再谈排名

本库原则：审「声称」必须回到声称的原件。先清点 PISA 到底是什么。

2.1 PISA 是什么：全球最大的 15 岁学生国际比较调查

OECD 官方对 PISA 的定义（PISA 官方主页）：

“PISA is the world’s largest international comparative survey of 15-year-old students, assessing the extent to which they have acquired key knowledge and skills essential for full participation in social and economic life.” [一手逐字]

PISA 2022 官方评估与分析框架（2023 年，全文取回）：

“The OECD Programme for International Student Assessment (PISA) assesses the extent to which 15-year-old students near the end of their compulsory education have acquired the knowledge and skills that are essential for full participation in modern societies. The assessment does not just ascertain whether students can reproduce knowledge; it also examines how well students can extrapolate from what they have learned and can apply that knowledge in unfamiliar settings, both in and outside of school.” [一手逐字]

关键事实（PISA 2022 官方结果第一卷「What is PISA」章，全文取回）：

  • PISA 是三年一轮的调查，1997 年启动、2000 年首次实施，2022 年是第八轮（原定 2021 年，因 COVID 推迟一年）
  • 测的是 15 岁学生（15 岁 3 个月至 16 岁 2 个月，已完成至少 6 年正式教育），关注阅读、数学、科学三个核心领域，每轮「主领域」轮换（2022 是数学）
  • 2022 年 81 个国家与经济体参与，约 69 万名学生参加，代表约 2,900 万 15 岁学生（官方逐字）
  • 2022 年新增创新领域「创造性思维」（creative thinking）

2.2 PISA 是抽样调查，不是普查

关键事实（PISA 2022 官方技术报告、结果第一卷 What is PISA 章）：

  • PISA 是基于样本的估计，不是普查。每个国家从目标总体（在校 15 岁学生）中抽学校、再抽学生
  • 所有估计都带标准误差（sampling error 来自抽样、measurement error 来自有限题目）
  • 官方用 95% 置信区间表述估计的不确定性（PISA 2022 结果第一卷附录 A3）：

“The statistics in this report represent estimates based on samples of students, rather than values that could be calculated if every student in every country had answered every question. Consequently, it is important to measure the degree of uncertainty of the estimates. In PISA, each estimate has an associated degree of uncertainty, which is expressed through a standard error.” [一手逐字]

2.3 背景问卷：PISA 测的不只是分数

PISA 同时收集学生/校长/家长/教师问卷，官方称其提供三类产出（PISA 2022 框架）：

“basic indicators that provide a profile of the knowledge and skills of students; indicators derived from the questionnaires that show how such skills relate to various demographic, social, economic and educational variables; indicators on trends that show changes in outcomes and their distributions…” [一手逐字]

守真锚小结：PISA 是一场真实运转、规模巨大、有测量学方法论的抽样评估。它测的是「15 岁在校学生某次测试的作答」，不是「教育质量」本身——这个区分是后续所有跳的起点。

三、刻度跳（一）：PISA 官方自己说「排名不能精确确定」

本节回答第一件可判定的事：PISA 排名有没有精确可言。

3.1 官方逐字：small differences 不应被考虑

PISA 2022 官方结果第一卷第二章「排名」段（官方页面 与 PDF 全文 逐字一致）：

“Moreover, many countries and economies score at similar levels; small differences that are not statistically significant or practically meaningful should not be considered (see Box 1 in Reader’s Guide).” [一手逐字]

“Because mean-score estimates are derived from samples and are thus associated with statistical uncertainty, it is often not possible to determine an exact ranking for all countries and economies. However, it is possible to identify the range of possible rankings for the country’s or economy’s mean performance. This range of ranks can be wide, particularly for countries/economies whose mean scores are similar to those of many other countries/economies.” [一手逐字]

这是官方自己写的：第一，很多国家分数相近；第二，小差异（不显著或没有实际意义）不应被考虑；第三，常无法确定所有国家与经济的精确排名；第四，排名区间可以很宽。

3.2 排名区间（range of ranks）是怎么算的

PISA 2022 官方结果第一卷附录 A3（官方页面，逐字）：

“Range-of-rank estimates are computed based on mean and standard-error-of-the-mean estimates for each country/economy, and take into account multiple comparisons amongst countries and economies at similar levels of performance.” [一手逐字]

附录 A3 详细方法（PDF 全文，逐字）：

“An estimate of the rank of a country mean, across all country means, can be derived from the estimates of the country means from student samples. However, because mean estimates have some degree of uncertainty, this uncertainty should also be reflected in the estimate of the rank. While mean estimates from samples follow a normal distribution, this is not the case of the rank estimates derived from these. Therefore, in order to construct a confidence interval for ranks, simulation methods were used. Data are simulated assuming that alternative mean estimates for each relevant country follow a normal distribution around the estimated mean, with a standard deviation equal to the standard error of the mean. Some 1 000 simulations are carried out…” [一手逐字]

也就是说：官方用 1,000 次蒙特卡洛模拟，给每个国家生成一个排名区间（如「第 8–15 名」），而不是一个精确名次。这是官方口径与本篇「刻度跳」的直接证据：官方给的是区间，媒体与政策读的是单点。

3.3 实证：美国 2022 数学的官方比较口径

美国教育部 NCES 官方指标页（International Comparisons, NCES，官方数据）：

  • 美国 2022 数学素养平均分 465 分，与 OECD 平均分「无显著差异」（not measurably different）
  • 与其余 80 个教育系统相比：43 个高于美国、25 个低于美国、12 个无显著差异（官方逐字）

“Compared with the 80 other education systems in PISA 2022, the U.S. average mathematics literacy score was higher than the average in 43 education systems; lower than the average in 25 education systems; and not measurably different from the average in 12 education systems.” [一手逐字]

注意官方用词：不是「第 28 名」这种单点，而是三组「显著高于/低于/无差异」。媒体头条用的「第 28 名」是媒体自己算的，官方明确给的是「与 43 个系统差异显著、12 个系统无差异」的区间式比较。

刻度跳小结：PISA 官方自己写了三段话——「小差异不应被考虑」「常无法确定精确排名」「排名区间可以很宽」——而媒体与政策最常用的恰恰是精确排名。这就是把「区间」读成「单点」的刻度跳。

四、刻度跳（二）：分数差与「一年学习」的官方换算

第二节回答第二件可判定的事：PISA 分数差能不能换算成「失去多少学习」。

4.1 官方换算：2022 数学下降 15 分 = 四分之三学年

PISA 2022 官方结果第一卷摘要（官方 PDF，逐字）：

“At the same time, on average, the PISA 2022 assessment saw an unprecedented drop in performance across the OECD. Compared to 2018, mean performance fell by ten score points in reading and by almost 15 score points in mathematics, which is equivalent to three-quarters of a year’s worth of learning. The decline in mathematics performance is three times greater than any previous consecutive change.” [一手逐字]

“In fact, one in four 15-year-old is now considered a low performer in mathematics, reading, and science on average across OECD countries.” [一手逐字]

官方自己做了这个换算：15 分 = 四分之三学年的学习。这是「刻度跳」的另一面——官方在报告里给出换算，媒体在标题里引用换算，政策在报告里引用换算。换算本身是官方给的，但换算的适用边界（见 4.2）官方也写了。

4.2 官方对换算的谨慎：PISA 2018 官方报告「分数是什么意思」

PISA 2018 官方结果第一卷「什么是 PISA 分数」章（官方页面，逐字）：

“PISA scores do not have a substantive meaning as they are not physical units, such as metres or grams. Instead, they are set in relation to the variation in results observed across all test participants. There is theoretically no minimum or maximum score in PISA; rather, the results are scaled to fit approximately normal distributions, with means around 500 score points and standard deviations around 100 score points. In statistical terms, a one-point difference on the PISA scale therefore corresponds to an effect size (Cohen’s d) of 0.01; and a 10-point difference to an effect size of 0.10.” [一手逐字]

关于「一年学习」换算的谨慎（同页，逐字）：

“There is considerable uncertainty about how PISA score-point differences translate into a metric such as ‘years of schooling’, and the empirical evidence is limited to a few countries and subjects.” [一手逐字]

“Because of the limited evidence about differences in PISA scores across school grades, for the same (or otherwise similar) students, and of the variability in these differences that is expected across subjects and countries, this report refrains from expressing PISA score differences in terms of an exact ‘years-of-schooling’ equivalent.” [一手逐字]

官方自相矛盾的两面：2022 版报告把 15 分换算成「四分之三学年」，2018 版报告却明确「refrains from expressing PISA score differences in terms of an exact ‘years-of-schooling’ equivalent」。官方在媒体压力下用了换算，又在方法论层面拒绝给出精确等价——这个张力本身就是刻度跳的活体。

4.3 OECD 官方研究：一年学习约 20 分

OECD 官方论文《How much do 15-year-olds learn over one year of schooling?》（Avvisati & Givord 2021，官方 PDF，逐字）：

“On average, what students learn over a school year corresponds to about 20 score points in PISA.” [一手逐字]

“students’ test scores increase by about one-fifth of a standard deviation over a ‘normal’ school year (or about 20 score points in PISA).” [一手逐字]

配套论文（EDU/WKP(2021)6，官方 PDF6/en/pdf)）：

“This study shows that in both Austria and Scotland (United Kingdom), students’ yearly learning progress around the age of 15 is equivalent to about one-fourth of a standard deviation in students’ test scores.” [一手逐字]

注意两组数字的差异：20 分（Avvisati & Givord 2021 国际比较）对 25 分（奥地利/苏格兰单国估计）对 25–30 分（Woessmann 2016 经验法则，PISA 2018 官方报告脚注转引）。官方自己承认这是「a broad generalisation, without taking it too literally」（PISA 2018 报告脚注引 Woessmann）。换算有多个版本、且官方自己说「不要太当真」——这是刻度跳的又一证据。

刻度跳小结：官方确实提供「15 分 = 四分之三学年」的换算，但官方同一文件里也承认「无法给出精确的 years-of-schooling 等价」、换算版本从 20 到 30 分不等、且经验法则「不要照字面理解」。媒体与政策只取了换算。分数差实质是「一组抽样估计的差异」，被读成了「精确的学习损失量」。

五、归属跳：增值模型（VAM）把「分数变化」归因给「教师」

本节回答第四件可判定的事：增值模型能不能把「学生分数变化」归因给「教师」。

5.1 什么是增值模型：定义与官方立场

美国统计学会（ASA）2014 年官方声明（官方 PDF，全文取回，逐字）：

“Many states and school districts have adopted Value-Added Models (VAMs) as part of educational accountability systems. The goal of these models, which are also referred to as Value-Added Assessment (VAA) Models, is to estimate effects of individual teachers or schools on student achievement while accounting for differences in student background. VAMs are increasingly promoted or mandated as a component in high-stakes decisions such as determining compensation, evaluating and ranking teachers, hiring or dismissing teachers, awarding tenure, and closing schools.” [一手逐字]

归属跳的起点：VAM 的目标是「estimate effects of individual teachers… on student achievement」——把「教师对成绩的影响」估计出来。这是把「学生分数的变化」归因给「教师」的尝试。

5.2 ASA 官方声明：相关不是因果，教师只占 1%–14%

ASA 2014 声明（官方 PDF，逐字）：

“VAMs typically measure correlation, not causation: Effects – positive or negative – attributed to a teacher may actually be caused by other factors that are not captured in the model.” [一手逐字]

“Most VAM studies find that teachers account for about 1% to 14% of the variability in test scores, and that the majority of opportunities for quality improvement are found in the system-level conditions.” [一手逐字]

“Ranking teachers by their VAM scores can have unintended consequences that reduce quality.” [一手逐字]

这是「归属跳」最直接的官方证据：统计学家的官方组织声明——VAM 测的是相关不是因果；教师只占成绩变异 1%–14%；按 VAM 排名可能有降低质量的副作用。ASA 还罕见地就一个政策议题公开发声（2014-04-08，官方新闻稿），其背景正是 VAM 被用于解雇、奖金、执照等高风险决策。

5.3 ASA 声明的背景：VAM 被用于高风险决策

ASA 2014 声明正文（官方 PDF，逐字）：

“The use of VAMs has grown over the last several years under the Obama administration’s Race to the Top initiative.” [多源交叉]

ASA 声明还提到一个具体案例（官方 PDF，逐字）：

“The Washington Post reported that scores in Washington, DC, Public Schools (DCPS) ‘IMPACT’ system, which evaluated teachers, had been erroneously calculated for 44 teachers, about 10% of those who had been evaluated through the DCPS VAM approach. One of those teachers was fired as a result of the error. (That teacher was rehired when the error was discovered.)” [一手逐字]

44 名教师、约 10%、其中一人被解雇后因错误发现而复职——这是 VAM 用于高风险决策出错的具体案例，写在统计学家的官方声明里。

5.4 年度稳定性：同一教师的 VAM 分数跨年相关只有 0.2–0.6

VAM 的「刻度」问题——同一教师今年的分数与明年的分数相关多高？多项研究：

  • McCaffrey, Sass, Lockwood & Mihaly (2009)（《Education Finance and Policy》4(4):572，DOI（出版商反爬，摘要级，正文经 eScholarship 全文 交叉核对））：

“We find year-to-year correlations in value-added measures in the range of 0.2–0.5 for elementary school and 0.3–0.7 for middle school teachers. Much of the variation in measured teacher performance (roughly 30–60 percent) is due to sampling error from ‘noise’ in student test scores.” [摘要逐字]

  • Kersting, Chen & Stigler (2013)（《Education Policy Analysis Archives》21(7)，DOI）：

“report year-to-year VAE correlations ranging from 0.2 to 0.6, indicating that in some cases rank-ordering of teachers changed substantially” [摘要逐字]

“between one quarter to one half of the teachers in the top quintile or quartile in one year remained in their group also the following year” [摘要逐字]

  • Koedel, Mihaly & Rockoff (2015) 综述（《Economics of Education Review》47:180-195，经 Span 综述 转引）：

“the correlation between a teacher’s value-added scores in 2 consecutive years; estimates from several studies range from 0.20 to 0.65” [多源交叉]

归属跳的刻度问题：同一教师跨年相关 0.2–0.6，意味着一年级的高分教师第二年可能跌入低分组——只有四分之一到一半的顶尖教师次年仍留在顶尖组。若把 VAM 分数当「教师质量的精确刻度」用于解雇，这就是把「有噪声的估计」当「精确的判据」。

5.5 Rothstein 反证：5 年级教师对 4 年级成绩有「显著影响」

Jesse Rothstein 的两篇论文（2009《Education Finance and Policy》4(4):537 与 2010《QJE》125(1):175）给出了 VAM 的证伪检验——用「未来教师不可能影响过去成绩」来检验 VAM 是否真的在测因果。

Rothstein 2009（Princeton 全文，逐字）：

“Non-random assignment of students to teachers can bias value added estimates of teachers’ causal effects. Rothstein (2008) shows that typical value added models indicate large counter-factual effects of 5th grade teachers on students’ 4th grade learning, implying that assignments do not satisfy the imposed assumptions.” [一手逐字]

Rothstein 2010（QJE，摘要，逐字）：

“In this paper, I develop falsification tests for three widely used VAM specifications, based on the idea that future teachers cannot influence students’ past achievement. In data from North Carolina, each of the VAMs’ exclusion restrictions is dramatically violated. In particular, these models indicate large ‘effects’ of fifth grade teachers on fourth grade test score gains.” [摘要逐字]

“I also find that conventional measures of individual teachers’ value added fade out very quickly and are at best weakly related to long-run effects. I discuss implications for the use of VAMs as personnel tools.” [摘要逐字]

这是「归属跳」最锋利的一击：如果 VAM 真的在测「教师贡献」，那么 5 年级教师对 4 年级成绩就不该有影响。但 VAM 显示「5 年级教师对 4 年级成绩有大影响」——这说明 VAM 里混入了学生分班的非随机性（学生被按能力分配），不是「贡献」。Rothstein 的结论是「VAMs need further development and validation before they can support causal interpretations or policy applications」（2009 全文逐字）。

5.6 归属跳的正面证据：Chetty-Friedman-Rockoff 2014

必须给对称面：Chetty, Friedman & Rockoff 2014（《AER》104(9):2633-79，Opportunity Insights 全文，逐字）：

“Are teachers’ impacts on students’ test scores (‘value-added’) a good measure of their quality? This question has sparked debate partly because of a lack of evidence on whether high value-added (VA) teachers improve students’ long-term outcomes. Using school district and tax records for more than one million children, we find that students assigned to high-VA teachers are more likely to attend college, earn higher salaries, and are less likely to have children as teenagers.” [一手逐字]

“Replacing a teacher whose VA is in the bottom 5% with an average teacher would increase the present value of students’ lifetime income by approximately $250,000 per classroom.” [一手逐字]

Chetty 用自己的数据（纽约州 3-8 年级、1989-2009，税务记录匹配约 100 万人）论证：VAM 确实能测出教师对长期结果的因果影响——高 VA 教师的学生更可能上大学、收入更高、更少青少年怀孕。其中「1 SD 教师 VA 提升使 28 岁收入 +1.3%」「大学入学 +0.82 个百分点」等数字（全文，逐字）。

Chetty 与 Rothstein 的正面交锋：Chetty 用「准实验」（教师离职带来的学生队列变化）交叉验证，认为自己的 VAM 无偏；Rothstein 用「未来教师不能影响过去」的证伪检验，认为 VAM 有偏。两人都对自己数据的特定设定成立，但结论相反——这正是「归属跳」的核心：VAM 能否归因教师，取决于数据、模型设定与检验方法，不是「VAM 能」或「VAM 不能」的简单答案。

5.7 归属跳小结

  • ASA（统计学家的官方组织）：相关不是因果，教师只占变异 1%–14%，按 VAM 排名可能降低质量
  • McCaffrey/Kersting/Koedel：VAM 跨年相关 0.2–0.6，顶尖组次年只有四分之一到一半留在组内
  • Rothstein：VAM 显示「5 年级教师影响 4 年级成绩」，说明分班非随机性混入，VAM 不能支持因果解释
  • Chetty：用准实验交叉验证，认为受控滞后分数的 VAM 无偏，且与长期收入相关

归属跳的结论：把「学生分数变化」读成「教师贡献」，是读一个有噪声、有偏、取决于模型设定的估计——ASA 说「相关不是因果」，Rothstein 说「模型有偏」，Chetty 说「特定设定下无偏」。官方统计学组织自己的结论是「应始终伴随精度度量与假设讨论」，而政策把它当精确判据用。

六、名号跳：NAEP 的「Proficient」与「报告卡」的名号

本节回答第五件可判定的事：NAEP 的「Proficient」到底是什么意思。

6.1 NAEP 是什么：美国唯一的全国性持续评估

NAGB（国家评估治理委员会）官方（About NAEP，逐字）：

“The National Assessment of Educational Progress (NAEP) is the largest nationally representative, continuing evaluation of the condition of education in the United States. It has served as a national yardstick of student achievement since 1969. Through The Nation’s Report Card, NAEP informs the public about what American students know and can do in various subject areas and compares achievement among states, large urban districts, and various student groups.” [一手逐字]

NCES 官方（The Nation’s Report Card，逐字）：

“NAEP. The National Assessment of Educational Progress (NAEP), also known as ‘the Nation’s Report Card,’ is the only nationally representative and continuing assessment of what America’s students know and can do in various subject areas.” [一手逐字]

6.2 名号跳：NAEP Proficient ≠ 年级水平

NAEP 官方网站（NAEP Achievement Levels，逐字）：

“Note: NAEP Proficient does not signify meeting grade level expectations, which are set through state assessments.” [一手逐字]

NAGB 官方（Achievement Levels 描述 PDF，逐字）：

“The NAEP Proficient level is intended to reflect solid academic performance and is not intended to match the proficiency levels set by State departments of education. Additionally, it does not signify ‘being on grade level.'” [一手逐字]

NAGB 官方《Closer Look at NAEP》（官方 PDF，逐字）：

“Myth: The NAEP Proficient level is like being on grade level. Fact: Proficient on NAEP means competency over challenging subject matter. This is not the same as being ‘on grade level,’ which refers to performance on local curriculum and standards.” [一手逐字]

名号跳的活体：NAEP 官方反复澄清「Proficient ≠ 年级水平」——这是官方「神话/事实」对照表里专门辟谣的一条。但媒体与政策长期把「NAEP Proficient 以下」读成「低于年级水平」。官方自己写了「Myth/Fact」，说明误读是普遍的、且官方在对抗它。

6.3 名号跳的历史：NAEP 标准被批评「太高」

NAEP 的 Proficient 标准从 1990 年设立起就争议不断（EdWeek 2018 报道，多源交叉）：

“The first evaluation—of the 1990 mathematics assessment—highlighted a number of serious concerns, and ultimately, this standard setting was redone.” [一手逐字]

  • 2002 年 NCLB 之后，各州用「proficient」命名自己的测试，与 NAEP 的含义混淆（EdWeek）
  • 2018 年 NAGB 修改措辞，把「grade」字样删掉、改为「NAEP proficient」（EdWeek）
  • 批评者（Superintendents Roundtable）称「多数国家的多数学生都过不了 NAEP 的 proficient 线」（EdWeek）

名号跳小结：NAEP 官方自己用「Myth/Fact」辟谣「Proficient 是年级水平」，并从 2018 年起把成绩描述里的「年级」字样删掉。但「报告卡」这个名号本身（Nation’s Report Card）就把「评估」翻译成了「报告卡」——一个让公众读出「成绩单」含义的隐喻。

七、消费账：教育测量在真实制度里怎么被用

本节回答第六件可判定的事：PISA/NAEP/VAM 在真实制度里怎么被用。

7.1 VAM 写入教师评估法规：从 Race to the Top 到州法

美国各州把 VAM 写入教师评估法规的规模（多个官方来源）：

“The number of states where evaluations are tied to tenure grew from none in 2009 to 19 in 2013.” [一手逐字]

“State law requires teachers to obtain an overall evaluation score of 4 or 5 during the last two years of a probationary period to be eligible for tenure.” [一手逐字]

“In 2011, the General Assembly expanded the definition of ‘inefficiency’ to include low evaluation scores (level 1 or 2).” [一手逐字]

  • 各州 VAM 使用比例（田纳西州报告，逐字）：田纳西 35%、佛罗里达 50%、俄亥俄 42.5%、DC 35%、北卡 16.7%——都是「学生成绩占教师评估的权重」[一手逐字]
  • 全国概览（Amrein-Beardsley《Putting Growth and Value-Added Models on the Map》，PDF，逐字）：

“Currently, 40 states and D.C. (80%) are using, piloting, or developing, some type of growth or value-added model.” [一手逐字]

“Ten states and D.C. (22%) tie (or are planning to tie) teacher tenure decisions to such output. Related, nine states and D.C. (20%) use (or are planning to use) these data to make teacher termination decisions.” [一手逐字]

“Six districts used or planned to use teacher evaluation scores to redeploy or release low performing teachers but narrowly defined the circumstances under which this could happen.” [一手逐字]

消费账核心：VAM 从 2009 年 Race to the Top 起被快速写入州法——到 2015 年 40 个州+DC 在使用或开发某种增值/成长模型，10 个州+DC 把 VAM 与解雇挂钩，19 个州 2013 年把评估与任期挂钩（2009 年为零）。ASA 2014 的声明正是针对这个「快速写入法规」的进程发出的。

7.2 PISA 冲击：国家排名的政策消费

PISA 排名被各国政策消费的经典案例：

  • 德国 PISA 冲击 2000：德国 2000 年首轮 PISA 排名不佳，引发「PISA-Schock」教育改革（Zhao 2020 综述，多源交叉）
  • 中国的「掐尖」批评：中国 B-S-J-Z（北京上海江苏浙江）2018 年排名第一，但被批评是「从最富裕地区挑选样本」（The 74 报道，多源交叉；Carnoy 2015《NEPC 简报》PDF，逐字）：

“The OECD has repeatedly held up Shanghai students and the Shanghai educational system as a model for the rest of the world and as representative of China, yet the sample is not representative even of the Shanghai 15-year-old population and certainly not of China.” [一手逐字]

“PISA’s director Andreas Schleicher ultimately admitted before a British House of Commons Education Select Committee that the Shanghai sample only represented 73% of that province’s 15-year-olds—this after a year of denying any problems with the Shanghai results.” [一手逐字]

  • PISA 2022 中国缺位：B-S-J-Z 参与 2022 轮但未报告成绩（官方逐字：学校在数据收集期关闭，PISA 2022 官方结果第一卷）：

“Chinese provinces/municipalities (Beijing, Shanghai, Jiangsu and Zhejiang) and Lebanon are participants in PISA 2022 but were unable to collect data because schools were closed during the intended data collection period.” [一手逐字]

消费账的活体：中国 2018 年排名第一被当「中国教育质量最高」的证据，但 2022 年 B-S-J-Z 不报成绩——同一个国家，一轮是「世界第一」，一轮是「缺席」。名号消费完全不看抽样条件。

7.3 teaching to the test 与作弊：Campbell 定律的实证

“The more any quantitative social indicator is used for social decision making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor.” [一手逐字]

  • Jacob & Levitt 2003（《QJE》118(3):843-877，NBER 全文，逐字）：

“We develop an algorithm for detecting teacher cheating that combines information on unexpected test score fluctuations and suspicious patterns of answers for students in a classroom.” [一手逐字]

“cheating occurs in 3-5 percent of elementary school classrooms each year in the Chicago Public Schools (CPS).” [一手逐字]

“Students in classrooms with large test-score gains that do not have suspicious answer string patterns retain the majority of their gains the next year, despite some loss due to mean reversion.” [一手逐字]

  • Jacob 2007（NBER WP 12817，NBER 页面，逐字）：

“This paper explores the phenomenon referred to as test score inflation, which occurs when achievement gains on ‘high-stakes’ exams outpace improvements on ‘low-stakes’ tests.” [一手逐字]

“I find evidence of considerable test score inflation in several different states, including those with quite different state testing systems.” [摘要逐字]

“We show that NCLB-era state tests predictably emphasized some state standards while consistently excluding others; a small number of standards typically accounted for a substantial fraction of test points. We find that students performed better on items testing frequently assessed standards—those that composed a larger fraction of the state test in prior years—which suggests that teachers targeted their instruction towards these predictably tested skills.” [摘要逐字]

消费账小结：高利害测试会诱导「教学重定向」（teaching to the test）与作弊（Chicago 3-5% 的班级每年有作弊迹象），state 测试的分差与 NAEP 分差出现「test score inflation」（应试膨胀）。Campbell 定律——指标用于决策越多，越容易被腐蚀——是教育测量消费账的总纲。

八、反虚无账：教育测量不是骗局

本节回答第七件可判定的事：教育测量是不是有真价值的制度。

8.1 PISA 有真价值：横向提供政策证据

教育部与学术界的正面评价（Ruiz-Primo et al. 2016，逐字）：

“PISA has served to build capacity and technical savvy in countries that do not have their own national assessments or the infrastructure to develop one. It has also served as an important yardstick to help countries understand aspects of their educational system. In these and other regards, PISA is a valuable instrument.” [一手逐字]

这篇论文（Ruiz-Primo 等，AERA 官方期刊《Educational Researcher》）是主张「对 PISA 结果要谨慎」的批评性论文，但哪怕批评者也说 PISA 是「valuable instrument」——反虚无账的对称面：批评者不否认 PISA 的价值。

8.2 NAEP 有真价值：纵向趋势是真实的

NAEP 的纵向趋势（自 1969 年起持续测量）是教育测量最被认可的价值之一。NCES 官方（About NAEP，逐字）：

“It has served as a national yardstick of student achievement since 1969.” [一手逐字]

NAEP 官方 FAQ（NCES FAQ，逐字）：

“Since 1969, National Assessment of Educational Progress (NAEP) has informed the public about what elementary and secondary students in the United States know and can do in various subject areas.” [一手逐字]

8.3 Chetty 的正面证据：VAM 有真实信号

Chetty 2014 证明 VAM 与长期结果（大学、收入、青少年怀孕）相关（全文，逐字）：

“We find that teacher VA has substantial impacts on a broad range of outcomes. A 1 SD improvement in teacher VA in a single grade raises the probability of college attendance at age 20 by 0.82 percentage points, relative to a sample mean of 37%.” [一手逐字]

“At age 28, the oldest age at which we currently have a sufficiently large sample size to estimate earnings impacts, a 1 SD increase in teacher quality in a single grade raises annual earnings by 1.3%.” [一手逐字]

注意：Chetty 自己也承认 VAM 有噪声（”because VA is estimated with noise, the gains from deselecting teachers based on data from a limited number of classrooms are smaller”），但他用「1 SD 教师 VA → 收入 +1.3%」证明 VAM 有真实信号。VAM 有噪声 ≠ VAM 无信号——这是反虚无账的核心。

8.4 效度传统：教育测量学的答案

教育测量学（psychometrics）对待「分数≠能力」的答案不是弃考，而是效度论证（Messick 框架、AERA/APA/NCME 标准）。AI 评测榜篇已详述（该篇 9.2 节，Messick 1995 逐字「validity is nothing less than an evaluative summary of both the evidence for and the actual as well as potential consequences of score interpretation and use」），本篇不重做，只做交叉引用。

反虚无账小结：批评者（Ruiz-Primo）说 PISA 是「valuable instrument」；NAEP 有 57 年纵向数据；Chetty 证明 VAM 有真实信号。「教育测量有缺陷」≠「教育测量是骗局」——缺陷是误差与误读，不是造假。

九、边界账与诚实空位

9.1 本篇不裁的事

  • 不裁任何国家、学校或教师个体的真实水平
  • 不裁「该不该排名」「该不该打分」的政策建议
  • 不重做 AI 评测篇已详述的 Messick 效度框架、ETS 考试安全、Popham teaching-to-the-test 区分
  • 不裁增值模型与班级规模/教师认证等替代政策孰优孰劣
  • 不裁中国教育质量、美国教育质量或任何国家的教育真实水平

9.2 诚实空位

  1. 美国 2022 数学排名区间（rank range）的具体数值：PISA 2022 官方结果第一卷的排名表（Table I.2.4）以图像形式存在于 PDF，pdftotext 无法提取；美国官方口径（NCES）给出的是「43 高/25 低/12 无差异」的三组比较，而非具体名次区间。本库未取到「第 8–15 名」这种具体区间数字，正文只采用 NCES 官方口径。
  2. PISA 官方 FAQ 排名问题的逐字回答：oecd.org 的 FAQ 页面有 Cloudflare 反爬，正文只拿到问题标题；PISA 2022 官方结果第一卷的「排名」段落（可逐字引用的权威表述）已完整取回，作为替代。
  3. 「PISA 冲击」的量化因果：德国 2000 年 PISA 冲击引发改革是媒体报道与学术共识，但「冲击导致的具体政策效果」缺少单一权威量化来源，本库按多源交叉标注。
  4. VAM 年度稳定性的完整元分析：Koedel-Mihaly-Rockoff 2015 综述的跨年相关 0.20–0.65 是多个研究的汇总，本库经 Springer 综述转引标注，未逐篇取回全部原始研究。
  5. 中国 B-S-J-Z 2022 不报成绩的官方原因：OECD 官方称「学校在数据收集期关闭」，但未解释为何其他地区（如香港、澳门）能报而 B-S-J-Z 不能；NBER 学者 Eberstadt 2026 年推测可能是「不愿公布可能下降的成绩」，这是推测、非官方口径，本库如实并列。

9.3 责任边界

  • 本篇是机制裁决，不构成对任何国家教育政策、任何学校或任何教师的评价
  • 不构成对任何考试（PISA/NAEP/TIMSS）或任何评估模型（VAM）的全面否定
  • 教育测量是复杂的专业领域，本库只做「承重」审计，不替代教育测量学专业判断

十、母题收口：被读成判断的那个数

教育测量是真实运转的制度——PISA 覆盖 81 国、NAEP 持续 57 年、VAM 写入数十州法规。 但「国家教育质量」与「教师优劣」这两个名号，从「分数」里读出来时，经历了三跳：

  1. 刻度跳：PISA 官方写「常无法确定精确排名」「排名区间可以很宽」「小差异不应被考虑」，媒体与政策读「第 28 名」；官方拿 15 分换算「四分之三学年」，同一官方文件里又承认「无法给出精确的 years-of-schooling 等价」。
  2. 归属跳：ASA 官方写「相关不是因果」「教师只占变异 1%–14%」「按 VAM 排名可能降低质量」，Rothstein 证明「5 年级教师对 4 年级成绩有显著影响」说明分班偏性混入，政策把 VAM 当解雇判据直接写入州法。
  3. 名号跳：NAEP 官方反复辟谣「Proficient ≠ 年级水平」，但「报告卡」的名号让公众读出「成绩单」；PISA 中国 2018 年「世界第一」2022 年「缺席」，名号消费完全不看抽样条件。

灵魂句：教育测量的三把尺都在量——量的是「抽样测试的作答」；被读成的是「一个国家的未来」与「一位教师的生死」。尺子没错，错的是把区间的中位数读成精确值、把相关读成贡献、把一次作答读成整个系统。

对称金句：官方说「it is often not possible to determine an exact ranking for all countries and economies」（PISA 2022 官方卷）——而全球媒体每轮都在发布精确排名；官方说「VAMs typically measure correlation, not causation」（ASA 2014）——而数十州把 VAM 写入解雇法规。两句都出自官方，读的人只取了对自己有利的那句。

来源清单

一手统筹（✓ 全文取回 / ◐ 摘要级 / ○ 多源交叉 / ⚠ 需读者亲核）

  1. ✓ OECD《PISA 2022 Results Volume I: The State of Learning and Equity in Education》（2023-12-05，全文 24MB PDF）：https://www.oecd.org/content/dam/oecd/en/publications/reports/2023/12/pisa-2022-results-volume-i_76772a36/53f23881-en.pdf
  2. ✓ OECD PISA 官方主页：https://www.oecd.org/en/about/programmes/pisa.html
  3. ✓ OECD《PISA 2022 Assessment and Analytical Framework》（2023，全文）：https://www.oecd.org/content/dam/oecd/en/publications/reports/2023/08/pisa-2022-assessment-and-analytical-framework_a124aec8/dfe0bf9c-en.pdf
  4. ✓ OECD《PISA 2022 Technical Report》（全文）：https://www.oecd.org/content/dam/oecd/en/publications/reports/2024/03/pisa-2022-technical-report_599753f0/01820d6d-en.pdf
  5. ✓ OECD《PISA 2022 结果第一卷「How did countries perform」章》（排名段逐字）：https://www.oecd.org/en/publications/2023/12/pisa-2022-results-volume-i_76772a36/full-report/how-did-countries-perform-in-pisa_dc514907.html
  6. ✓ OECD《PISA 2022 结果第一卷「What is PISA」章》（B-S-J-Z 段逐字）：https://www.oecd.org/en/publications/pisa-2022-results-volume-i_53f23881-en/full-report/what-is-pisa_b84730a6.html
  7. ✓ OECD《PISA 2022 得分与排名 FAQ》（页面反爬，标题级）：https://www.oecd.org/en/about/programmes/pisa/pisa-frequently-asked-questions-faqs.html
  8. ✓ OECD 附录 A3 技术注释（标准误差/置信区间/排名区间逐字）：https://www.oecd-ilibrary.org/sites/e5cd1bfd-en/index.html?itemId=/content/component/e5cd1bfd-en
  9. ✓ OECD《PISA 2018 Results Volume I「What is a PISA score」章》（分数无物理意义/换算谨慎逐字）：https://www.oecd.org/en/publications/pisa-2018-results-volume-i_5f07c754-en/full-report/component-8.html
  10. ✓ OECD《How much do 15-year-olds learn over one year of schooling?》（Avvisati & Givord 2021，约 20 分/年逐字）：https://www.oecd.org/content/dam/oecd/en/publications/reports/2021/10/how-much-do-15-year-olds-learn-over-one-year-of-schooling_bf040ad5/b837fd6a-en.pdf
  11. ✓ OECD《The learning gain over one school year》（EDU/WKP(2021)6，约 1/4 SD 逐字）：https://one.oecd.org/document/EDU/WKP(2021)6/en/pdf6/en/pdf)
  12. ✓ OECD《PISA 2022 Insights and Interpretations》（分数表）：https://www.oecd.org/content/dam/oecd/en/publications/support-materials/2023/12/pisa-2022-results-volume-i_76772a36/PISA%202022%20Insights%20and%20Interpretations.pdf
  13. ✓ OECD Education GPS（美国 PISA 2022 数据）：https://gpseducation.oecd.org/CountryProfile?primaryCountry=USA&topic=PI&treshold=10
  14. ✓ NCES《International Comparisons: Reading, Mathematics, and Science Literacy of 15-Year-Old Students》（美国 465 分/43 高 25 低 12 无差异逐字）：https://nces.ed.gov/programs/coe/indicator/cnu?tid=4
  15. ✓ NAGB《About NAEP》（1971 年起报告卡逐字）：https://www.nagb.gov/naep/about-naep.html
  16. ✓ NCES《The Nation’s Report Card》官方页：https://nces.ed.gov/nationsreportcard/
  17. ✓ NCES《Understanding Results》（Proficient ≠ 年级水平逐字）：https://nces.ed.gov/nationsreportcard/guides/
  18. ✓ NCES《NAEP Achievement Levels》（Proficient ≠ 年级水平逐字）：https://nces.ed.gov/nationsreportcard/about/achieve.aspx
  19. ✓ NAGB《Achievement Levels 描述》（Proficient ≠ 年级水平逐字）：https://www.nagb.gov/content/dam/nagb/en/documents/naep/achievement-levels-descriptions.pdf
  20. ✓ NAGB《A Closer Look at NAEP》（Myth/Fact 逐字）：https://www.nagb.gov/content/dam/nagb/en/documents/a-closer-look-at-naep.pdf
  21. ✓ NCES《NAEP FAQ》：https://nces.ed.gov/nationsreportcard/about/faqs.aspx
  22. ✓ ASA《ASA Statement on Using Value-Added Models for Educational Assessment》（2014-04，全文逐字）：https://www.amstat.org/asa/files/pdfs/POL-ASAVAM-Statement.pdf
  23. ✓ ASA 官方新闻稿（2014-04-08）：https://www.amstat.org/docs/default-source/amstat-documents/pol-asavam-statementpr.pdf
  24. ✓ ASA《ASA Statement on Value-Added Models》发表于《Journal of Educational and Behavioral Statistics》（Morganstein & Wasserstein 2014）：https://www.tandfonline.com/doi/full/10.1080/2330443X.2014.956906
  25. ✓ Chetty, Friedman & Rockoff 2014《Measuring the Impacts of Teachers II》《AER》104(9):2633-79（全文 PDF）：https://opportunityinsights.org/wp-content/uploads/2018/03/teachers2.pdf
  26. ✓ Chetty 2014 官方页面（摘要/数据）：https://opportunityinsights.org/paper/teachersii/
  27. ◐ Rothstein 2009《Student Sorting and Bias in Value-Added Estimation》《Education Finance and Policy》4(4):537-571（Princeton 全文）：https://www.princeton.edu/~ceps/workingpapers/170rothstein.pdf
  28. ◐ Rothstein 2010《Teacher Quality in Educational Production》《QJE》125(1):175-214（摘要逐字）：https://ideas.repec.org/a/oup/qjecon/v125y2010i1p175-214..html
  29. ◐ McCaffrey, Sass, Lockwood & Mihaly 2009《The Intertemporal Variability of Teacher Effect Estimates》（出版商反爬，摘要级，eScholarship 全文交叉核对）：https://doi.org/10.1162/edfp.2009.4.4.572（eScholarship）
  30. ✓ Kersting, Chen & Stigler 2013《Value-added Teacher Estimates as Part of Teacher Evaluations》《EPAA》21(7)：https://doi.org/10.14507/epaa.v21n7.2013
  31. ○ Koedel, Mihaly & Rockoff 2015《Value-added modeling: A review》《Economics of Education Review》47:180-195（经 Springer 综述转引）：https://link.springer.com/article/10.1007/s11092-019-09303-w
  32. ✓ Guan et al. 2019《Methodological issues in value-added modeling: an international review from 26 countries》（Springer 综述）：https://link.springer.com/article/10.1007/s11092-019-09303-w
  33. ✓ Ruiz-Primo et al. 2016《A Call for a More Measured Approach to Reporting and Interpreting PISA Results》《Educational Researcher》（批评者说 PISA 是 valuable instrument）：https://journals.sagepub.com/doi/10.3102/0013189X16649961
  34. ◐ Komatsu & Rappleye 2017《A new global policy regime founded on invalid statistics?》《Comparative Education》53(2)（Hanushek-Woessmann 增长声称无效）：https://www.tandfonline.com/doi/abs/10.1080/03050068.2017.1300008
  35. ✓ Carnoy 2015《国际测试分数的批评》NEPC 简报（上海样本/卫生问题逐字）：https://www.nepc.colorado.edu/sites/default/files/pb_carnoy_international_test_scores_0.pdf
  36. ✓ Zhao《The PISA Illusion》（2019-12-05，Komatsu-Rappleye 逻辑批评转述）：https://zhaolearning.org/2019/12/05/the-pisa-illusion/
  37. ✓ Jacob & Levitt 2003《Rotten Apples》《QJE》118(3):843-877（NBER 全文，3-5% 作弊逐字）：https://www.nber.org/system/files/working_papers/w9413/w9413.pdf
  38. ✓ Jacob & Levitt 2003《Catching Cheating Teachers》（NBER WP 9414）：https://www.nber.org/system/files/working_papers/w9414/w9414.pdf
  39. ✓ Jacob 2007《Test-Based Accountability and Student Achievement》（NBER WP 12817，test score inflation 逐字）：https://ideas.repec.org/p/nbr/nberwo/12817.html
  40. ✓ Wang / 其他 2013《Teaching to the Test in the NCLB Era》：https://journals.sagepub.com/doi/10.3102/0013189X14554449
  41. ✓ Berliner & Nichols《The Inevitable Corruption of Indicators and Educators Through High-Stakes Testing》（Campbell 定律逐字）：https://www.nepc.colorado.edu/sites/default/files/EPSL-0503-101-EPRU.pdf
  42. ✓ 田纳西州审计办公室《Use of Value-Added in Teacher Evaluations》（2015，VAM 权重/USDOE 延期逐字）：https://comptroller.tn.gov/content/dam/cot/orea/advanced-search/2015/2015_OREA_TVAASandTchrEval.pdf
  43. ✓ USDOE《Study of Emerging Teacher Evaluation Systems》（2016，VAM 使用/解雇逐字）：https://www.ed.gov/sites/ed/files/rschstat/eval/teaching/emerging-teacher-evaluation/report.pdf
  44. ✓ Amrein-Beardsley《Putting Growth and Value-Added Models on the Map》（40 州+DC 逐字）：https://www.nysed.gov/sites/default/files/beardsleyputtinggrowthandvalueaddedmodel.pdf
  45. ✓ NCTQ《State of the States 2015》（27 州年评/17 州 growth 主导/23 州 tenure 挂钩逐字）：https://www.teachingchannel.com/wp-content/uploads/2023/03/state-of-the-states-2015_lores.pdf
  46. ✓ ECS《State Education Policy》（2014，16 州 tenure 挂钩逐字）：http://www.ecs.org/clearinghouse/01/12/42/11242.pdf
  47. ✓ EdWeek《Is ‘Proficient’ Insufficient?》（2018，NAEP 标准争议多源）：https://www.edweek.org/teaching-learning/is-proficient-insufficient-a-new-wrinkle-in-the-debate-over-naep-achievement-levels/2018/11
  48. ✓ 国家科学院《Evaluation of the Achievement Levels for Mathematics and Reading on NAEP》（1990 标准重做逐字）：https://www.nationalacademies.org/read/23409/chapter/4
  49. ○ The 74《American Math Scores Fall on International Test》（美国 2022/中国 B-S-J-Z 缺位多源）：https://www.the74million.org/article/american-math-scores-fall-on-international-test-but-many-other-countries-suffered-more/
  50. ○ Eberstadt《Are China’s Students Really Number One?》（AEI/Milken Review，2026，推测口径）：https://www.aei.org/research-products/journal-publication/are-chinas-students-really-number-one-a-statistical-riddle/
  51. ✓ OECD《PISA Participants》（中国 B-S-J-Z 参与年份）：https://www.oecd.org/en/about/programmes/pisa/pisa-participants.html
  52. ✓ NCES《PISA Participation by Year》（B-S-J-G/B-S-J-Z 定义）：https://nces.ed.gov/surveys/pisa/participation.asp
  53. ✓ 英国教育部《What is PISA data and how does it measure students’ success at school?》（OSR 谨慎逐字）：https://educationhub.blog.gov.uk/2024/09/18/what-is-pisa-data-and-how-does-it-measure-students-success-at-school/
  54. ✓ Rockoff《Discussion of the ASA’s Statement on VAM》（Chetty 团队对 ASA 的回应）：http://www.equality-of-opportunity.org/assets/documents/teachers_discussion_asa.pdf

取不回清单（如实登记）

  • PISA 2022 官方结果第一卷排名表（Table I.2.4）的具体排名区间数值（PDF 表格为图像，pdftotext 无法提取；官方口径以 NCES 三组比较为准，不虚构具体名次）
  • PISA 官方 FAQ 排名问题的逐字回答（oecd.org Cloudflare 反爬，正文以官方结果卷的「排名」段替代）
  • 美国 2018 年数学「25 名/37 OECD」的官方出处（NSF 报告引述，非 NCES 一手，按多源标注）
  • 上海 2018 年样本「73% 代表」的英国议会听证记录原文（Carnoy 转述，未取回听证会原记录）

外链实测说明：本清单 54 条来源中 51 个唯一 URL 于 2026-08-03 实测，48×200/1×403（McCaffrey 2009 DOI 为 MIT Press 反爬，已配 eScholarship 全文替代通路）/其余为转引标注。oecd.org 与 nces.ed.gov 部分页面有反爬，但承重引用均通过官方 PDF 或 exa 全文抓取逐字核对。

标签: