📚 How the 2024 National Reference Test Results Shape GCSE Grading Standards | 2024全国参考测试结果如何影响GCSE评分基准
GCSE grade boundaries in English Language and Mathematics are not decided by guesswork. Since 2017, Ofqual has had an independent measuring instrument - the National Reference Test - designed to detect whether the national cohort of Year 11 students has genuinely become stronger or weaker in those two subjects. The 2024 sitting produced evidence that fed directly into the summer 2024 award. This article explains what the 2024 NRT measured, how the statistics are handled, and precisely how far it can move a grade boundary.
GCSE 英语语言与数学的分数线并非凭猜测决定。自 2017 年起，Ofqual 拥有一件独立的测量工具——全国参考测试（NRT）——用来判断全国 11 年级考生在这两门科目上是否真的变强或变弱。2024 年的施测结果直接进入了 2024 年夏季评分流程。本文将说明 2024 年 NRT 究竟测了什么、统计数据如何处理，以及它能在多大程度上改变一条分数线。
1. What the National Reference Test Actually Is | 全国参考测试究竟是什么
The National Reference Test is an England-only assessment commissioned by Ofqual, the qualifications regulator, and delivered by the National Foundation for Educational Research (NFER). It is sat each spring by a sample of Year 11 students in exactly two subjects: GCSE English Language and GCSE Mathematics.
全国参考测试是仅限英格兰的测评项目，由资格与考试监管机构 Ofqual 委托，由国家教育研究基金会（NFER）负责实施。每年春季，一批抽样选出的 11 年级学生参加该测试，科目只有两门：GCSE 英语语言与 GCSE 数学。
The test is not a qualification. Students receive no grade, no certificate and no individual feedback. The results exist only as statistical evidence about how the ability of the national cohort has changed over time.
该测试不是资格证书。学生不会获得成绩、证书或个人反馈。其结果只作为统计证据存在，用于说明全国考生群体的整体能力随时间发生了怎样的变化。
First sat in 2017, the NRT has built a stable baseline against which every later year can be compared. Roughly 300 schools and up to about 30,000 students take part in each subject annually.
NRT 于 2017 年首次施测，此后建立了稳定基准线，可用于与之后每一年进行比较。每门科目每年大约有 300 所学校、最多约 3 万名学生参与。
Crucially, the NRT papers are assembled from items that have previously appeared in real GCSE exams. That shared item bank is what makes the NRT a measuring stick for GCSE standards rather than an unrelated quiz.
关键在于，NRT 试卷由曾经出现在真实 GCSE 考试中的题目组成。正是这个共享题库，使 NRT 成为衡量 GCSE 标准的标尺，而不是一场与考试无关的测验。
2. Why Ofqual Needs an Independent Measuring Stick | 为什么 Ofqual 需要一把独立的标尺
The central problem in standards maintenance is that grade standards should stay constant while cohorts change. If the students sitting GCSEs in a given year are genuinely weaker than the year before, keeping the same grade boundaries would mean standards had fallen.
标准维持的核心难题在于：评分标准应保持恒定，而考生群体却在变化。如果某年参加 GCSE 的学生确实比上一年弱，却仍沿用同样的分数线，那就意味着标准下降了。
Ofqual's main statistical anchor is prior attainment - typically Key Stage 2 results from primary school. This supports the "comparable outcomes" approach: if the cohort's prior attainment is similar, results should be similar.
Ofqual 的主要统计锚点是先前的学业表现——通常是小学阶段的 Key Stage 2 成绩。这支撑了"可比结果"原则：如果考生群体的先期学业水平相近，那么整体结果应当相近。
Prior attainment has limits. It is historical, it reflects a different curriculum, and it does not measure how well Year 11 students actually perform in English and mathematics today. The NRT fills that gap with direct, current, subject-specific evidence.
但先期成绩有其局限：它是历史数据，反映的是另一套课程，也无法测量今天 11 年级学生在英语和数学上的真实表现。NRT 正是用直接的、当下的、科目专属的证据来填补这一空白。
In short, comparable outcomes tells Ofqual what to expect; the NRT tells Ofqual what actually happened.
简言之，"可比结果"告诉 Ofqual 应该预期什么，而 NRT 告诉 Ofqual 实际发生了什么。
3. Design and Sampling: How the 2024 NRT Was Run | 设计与抽样：2024 年 NRT 如何实施
Schools are selected at random but with stratification, so that the sample matches the national Year 11 population on key characteristics such as region, school type, school size and prior attainment profile. Schools that decline are replaced to protect representativeness.
学校是随机抽取的，但采用分层抽样，使样本在地域、学校类型、学校规模以及生源学业水平等关键特征上与全国 11 年级人口结构相匹配。拒绝参与的学校会被替换，以保证代表性。
Students sit one of the two subjects. English Language papers run for roughly one hour and five minutes; Mathematics papers run for roughly one hour. The test is administered by external, trained administrators under standardised conditions, which reduces the risk of schools "teaching to the test".
学生只参加其中一门科目。英语语言试卷约为 1 小时 5 分钟，数学试卷约为 1 小时。测试由经过培训的外部监考人员按统一条件实施，从而降低了学校"应试式教学"干扰结果的风险。
The 2024 sitting took place in the spring term, typically between late February and late March, before the GCSE examinations. Testing Year 11 students at this point captures cohort ability before the pressure of final revision distorts the picture.
2024 年的施测安排在第一学期的春季，通常在 2 月下旬至 3 月下旬之间，早于 GCSE 正式考试。在这一时点测试 11 年级学生，可以在应试复习压力扭曲结果之前捕捉到考生群体的真实能力。
Results are then weighted back to the national population. A well-designed sample of this size gives Ofqual a reliable national estimate with relatively narrow confidence intervals.
随后，结果会被加权还原到全国总体。如此规模且设计良好的样本，能够为 Ofqual 提供可靠且置信区间相对较窄的全国估计值。
4. The 2024 Results in Context | 2024 年结果及其背景
Ofqual receives the NRT analysis in the summer, in time to inform the awarding meetings for that year's GCSE English Language and Mathematics. The 2024 report was used in exactly that way for the August 2024 award.
Ofqual 在夏季收到 NRT 分析结果，正好赶上当年 GCSE 英语语言与数学的评分会议。2024 年的报告正是以这一方式用于 2024 年 8 月的评分工作。
The headline message of the 2024 NRT was stability rather than drama. Reported performance in both English Language and Mathematics was close to the 2023 level, and the year-on-year differences were small - in most cases falling within the margin of statistical error.
2024 年 NRT 传达的核心信息是稳定，而非剧变。英语语言与数学的报告表现都与 2023 年水平接近，年度差异很小，多数情况下落在统计误差范围之内。
Over the longer run, the pattern is worth noting. Mathematics performance has fluctuated around the 2017 baseline with no clear sustained trend, while English Language has drifted gently downward since the baseline year. This drift has been gradual year on year, not a sudden cliff edge.
从更长时间来看，这一趋势值得注意。数学表现围绕 2017 年基准线波动，没有明显的持续走向；而英语语言自基准年以来呈缓慢下滑。这种下滑是逐年渐进的，并非某一年突然断崖。
Because the 2024 changes were small and not statistically clear-cut, the evidence did not provide a basis for shifting standards in either subject. Readers wanting the exact published figures should consult Ofqual's official NRT report and NFER's technical documentation.
由于 2024 年的变化幅度小、在统计上并不明确，这些证据不足以支撑对任一科目进行标准调整。如需确切公布数值，读者应查阅 Ofqual 的官方 NRT 报告与 NFER 的技术文件。
5. From Raw Marks to a Common Scale: Rasch and IRT | 从原始分到共同量表：Rasch 与项目反应理论
You cannot simply compare average marks across years, because the test is not identical each year and the students are different people. The NRT solves this with item response modelling, specifically a Rasch model, which places both items and students on a single common scale.
不能简单比较各年度的平均分，因为每年试卷并不完全相同，考生也是不同的人。NRT 用项目反应模型（具体为 Rasch 模型）解决这一问题，把题目与学生放在同一个共同量表中。
The probability that a student answers an item correctly is modelled as a function of the difference between the student's ability and the item's difficulty:
学生答对某题的概率，被建模为学生能力与题目难度之差的函数：
P(correct) = eθ − b ÷ (1 + eθ − b)
Here θ (theta) is the student's ability in logits and b is the item difficulty in logits. Because a core set of anchor items is carried over from previous GCSE papers and previous NRT sittings, the scale itself stays fixed even when the test form changes.
其中 θ（theta）是学生的能力值（以 logit 为单位），b 是题目难度值（同样以 logit 为单位）。由于有一批核心锚题沿用自历年 GCSE 试卷与往届 NRT，即使试卷组成变化，量表本身也保持固定。
The output is therefore not "the 2024 cohort scored 27 out of 40". It is "the 2024 cohort's mean ability estimate is X logits, with a 95% confidence interval of Y to Z, compared with the 2017 baseline".
因此输出结果不是"2024 届考生 40 分中考了 27 分"，而是"
Published by TutorHao | Exam Prep Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图