過了十二年，
十二年國教入學改革
為學生帶來什麼，又犧牲什麼？ Twelve years on:
what did Taiwan's admission reform
give students, and what did it cost?

先看它想做什麼Start with what it set out to do

在問「做到了嗎」之前，得先知道它答應了什麼。教育部〈十二年國民基本教育實施計畫〉列了七項總體目標，其中第三項是這樣寫的：

Before asking whether it worked, we need to know what it promised. The Ministry of Education's implementation plan lists seven overall goals. The third reads:

促進教育機會均等，以實現社會公平與正義。

教育部〈十二年國民基本教育實施計畫（核定本）〉・總體目標第三項

“To promote equality of educational opportunity, so as to realise social fairness and justice.”

Ministry of Education, Implementation Plan for Twelve-Year Basic Education, overall goal 3

同一份計畫裡，還有幾句同樣清楚的話。五大理念的第一項是「有教無類」——「高級中等教育階段是以全體 15 歲以上的國民為對象，不分種族、性別、階級、社經條件、地區等，教育機會一律均等」。第五項「優質銜接」則說要「藉由高級中等學校的均優質化，均衡城鄉教育資源」。總體目標第六項還寫著要「有效舒緩過度升學壓力」。

The same document contains other equally explicit statements. The first of its five guiding principles is “education for all without distinction” — “at the senior secondary stage, all citizens aged 15 and over are the object, regardless of ethnicity, sex, class, socioeconomic condition or region; educational opportunity shall be equal for all.” The fifth, “quality articulation,” calls for “levelling up the quality of senior secondary schools so as to balance urban and rural educational resources.” Goal six adds “effectively easing excessive pressure to advance.”

先講結論：兩個做到了，一個代價The short answer: two goals met, one price paid

先說好消息，而且是兩個。第一，學校真的變得比較「均質」了——明星學校不再只收同一種學生，同一所高中裡的學生程度差距明顯拉開。第二，教育機會真的被重新分配了——成績後段的學生進到更好的學校、遇到更強的同儕、遇到學歷更高的老師。這兩件事就是計畫書上那句「促進教育機會均等」，而且它們真的發生了。

Start with the good news — there are two pieces of it. First, schools really did become more evenly mixed. Top-ranked schools no longer take only one kind of student, and the spread of achievement inside a given high school widened noticeably. Second, opportunity really was redistributed: lower-achieving students moved into better schools, with stronger classmates and better-credentialed teachers. Those two things are what “promote equality of educational opportunity” meant in practice — and they happened.

代價也是真的。同一時間，高中生的平均成績下降，而且從最前段到最後段的學生一起下降——沒有一群人因為這個改革在成績上受惠。

The cost is real too. Over the same period average performance fell, and it fell right across the distribution, from the top decile to the bottom. On test scores, no group came out ahead.

這個政策到底改了什麼？What the reform actually changed

2014 年以前，國三學生考「基本學力測驗」，分數幾乎決定了你能上哪一所高中。2014 年以後，改考「國中教育會考」，而且教育部規定：考試成績在入學積分裡最多只能占三分之一。其餘的部分由各就學區自己設計，通常包含多元學習表現（志工服務、競賽、獎懲紀錄等）與志願序。

Before 2014, ninth graders took the Basic Competence Test, and that score largely determined which high school they could attend. From 2014 the exam was replaced by the Comprehensive Assessment Program (CAP), and the Ministry of Education stipulated that test scores may account for no more than one third of a student's admission points. The rest is designed by each of the school districts, and typically includes “multi-literacy performance” (volunteering, competitions, school records) and the order in which a student ranks their preferred schools.

「分流」是什麼？為什麼它是重點What “tracking” means, and why it is the point

如果每一所高中收到的學生程度都差不多，那所學校裡的學生成績會有很大的高低差距。反過來，如果一所學校只收前 3% 的學生，那所學校裡的學生程度會非常整齊。所以有一個很直接的量尺：把一所學校裡學生成績的落差算出來，落差越小，代表學生越是被按照能力挑進來的。

If every high school received a mixed intake, the spread of achievement inside each school would be wide. If a school admits only the top 3%, the students inside it will look very similar to one another. That gives a direct yardstick: measure the spread of achievement inside a school. The smaller the spread, the more students were sorted in by ability.

怎麼知道是政策造成的，不是碰巧？How do we know it was the policy?

這是整個研究最關鍵的一步，也是最容易被忽略的一步。成績會因為很多原因變動——景氣、疫情、少子化、補習班、手機。光看「改革前後有沒有變」是不夠的。

This is the single most important step in the study, and the easiest one to skip past. Test scores move for all sorts of reasons — the economy, a pandemic, falling birth rates, cram schools, smartphones. Simply comparing before and after is not enough.

這篇研究用的方法叫做雙重差分（difference-in-differences）。想法很單純：找一群「同時代、同社會、但沒被這個政策掃到」的人當對照。這裡的對照組是國中生——他們和高中生活在同一個時代、經歷同樣的疫情與社會變化，但國中入學是按住家學區分發的，那套規則在這段期間沒有改。

The study uses difference-in-differences. The idea is simple: find a group living through the same era and the same society, but not touched by this particular policy. Here that group is ninth graders. They share the same decade, the same pandemic, the same social change as tenth graders — but junior high admission runs on residential catchment, and those rules did not change during the period.

資料來自經濟合作暨發展組織（OECD）的國際學生能力評量計畫（PISA）——每三年對 15 歲學生做的國際評量。15 歲剛好橫跨國三與高一，所以同一份資料裡同時有對照組和處理組。台灣參加了六次。

The data come from the Programme for International Student Assessment (PISA), the OECD's international assessment of 15-year-olds, run every three years. Fifteen is exactly the age that straddles ninth and tenth grade, so a single dataset contains both the control and the treated group. Taiwan took part six times.

做到了 ①：學校真的變得比較均質Delivered ①: schools really did even out

對應的是計畫書裡的「均優質化」與「均衡城鄉教育資源」。要達到這件事，前提是明星學校不能再壟斷同一批學生——所以我們先看校內程度差距有沒有拉開。

This corresponds to the plan's language about “levelling up school quality” and “balancing urban and rural educational resources.” For that to happen, top schools must stop monopolising the same pool of students — so the first thing to check is whether the spread inside schools widened.

校內的成績落差擴大了 30 分（效果量 0.86，以政策前高中的標準差為基準）。這不是小數字：它等於補回「考試分數說了算」與「按住家分發」之間大約一半的距離。三科分開看，數學、科學、閱讀全部同方向。

Within-school spread widened by 30 points (effect size 0.86, against the pre-reform high school baseline). That is not a small number: it closes about half the distance between “scores decide everything” and “your address decides everything.” Broken out by subject, maths, science and reading all move the same way.

做到了 ②：教育機會真的被重新分配Delivered ②: opportunity really was redistributed

這一節直接對應那句「促進教育機會均等，以實現社會公平與正義」。學校變均質只是手段，真正要問的是：原本進不去的孩子，有沒有真的拿到比較好的東西？

This section speaks directly to “promote equality of educational opportunity, so as to realise social fairness and justice.” Evening out schools is only the means. The real question is: did the students who would not have got in actually end up with something better?

如果分流鬆了，那麼原本被擋在門外的學生，應該真的走進了比較好的學校。研究把學生按成績分成四組（Q1 最低、Q4 最高），看他們接觸到的學校資源怎麼變。

If sorting loosened, then students previously kept outside the gate should now be inside better schools. The study splits students into achievement quartiles (Q1 lowest, Q4 highest) and asks what school resources each group ends up with.

代價：平均成績掉了，而且是一起掉The cost: scores fell, and fell together

高中生的 PISA 三科總分平均下降 50 分，相當於 0.18 個標準差。這裡最值得注意的不是那個數字有多大，而是它的形狀：研究把學生從最低 10% 排到最高 10%，逐段估計政策效果——每一段都是負的，而且彼此沒有顯著差異。

The PISA 3-subject total fell by 50 points on average, or 0.18 standard deviations. What matters most here is not the size of that number but its shape. The study estimates the effect separately at every point in the achievement distribution, from the bottom tenth to the top tenth. Every point is negative, and none is significantly different from any other.

所以，我們學到什麼？So what did we learn?

一個為了打散學校分層、讓入學機會更公平而設計的政策，確實做到了它想做的事；但它同時讓所有學生的平均表現下降。這是一個真實存在的取捨，不是誰做錯了什麼。

A policy designed to break up school stratification and widen access did exactly that. It also lowered average performance for everyone. This is a real trade-off, not somebody's mistake.

回到最前面那個問題——十二年國教有沒有做到「促進教育機會均等，以實現社會公平與正義」？就這篇研究能看到的範圍而言，答案是：有。學校之間的分層鬆開約一半，後段學生真的進到更好的學校、遇到更強的同儕與學歷更高的老師，而且沒有變成用家世取代分數。這不是宣傳，是可以量出來的。

Back to the question at the top — did the reform deliver “equality of educational opportunity, so as to realise social fairness and justice”? Within what this study can see, the answer is yes. Between-school sorting loosened by about half; lower-achieving students really did reach better schools, stronger peers and better-credentialed teachers; and background did not quietly replace test scores as the sorting device. That is measurable, not rhetoric.

但同一份資料也給了第二個答案：機會確實被重新分配了，可是機會沒有自動變成成績。後段學生進到了更好的學校、遇到更強的同儕與學歷更高的老師——這些都真實發生了——可是這些並沒有轉換成更高的分數。同時，教室裡學生程度變得更參差，教學的難度上升，於是所有人的表現一起被拉下來。

The same data gives a second answer: opportunity was redistributed, but opportunity did not automatically become achievement. Lower-achieving students did move into better schools, with stronger peers and better-credentialed teachers — all of that really happened — and none of it translated into higher scores. Meanwhile classrooms became more heterogeneous, teaching got harder, and performance fell for everyone.

這個研究不能回答的事What this study cannot tell you

一份誠實的研究要能說清楚自己的邊界。

An honest study should be able to state its own boundaries.

1. 測量的時間點太早。PISA 在四、五月施測，學生才進高中一個半學期。理想上，「校內落差」該在入學當下量，「學習成果」該在幾年後量。如果分流的影響會隨時間累積，這篇論文估到的效果是偏小的。
2. 它量的是「分流」，不是整個十二年國教。國中生也會被改革影響——他們可能因為考試變得不重要而少念一點書。但因為兩組人在國中階段受到的影響是一樣的，這部分在雙重差分裡被抵銷掉了。所以這篇論文識別的是高中端分流變化的淨效果，不是整包政策的總效果。
3. 它說不出成績為什麼掉。可能是教室變得太參差、老師難教；也可能是資源重分配與同儕效應。這篇論文無法分辨是哪一個，或者三個都有。

1. The timing of measurement is too early. PISA is administered in April or May, about one and a half semesters into high school. Ideally, within-school spread would be measured at enrolment and achievement measured years later. If tracking effects accumulate over time, this study's estimates are too small.
2. It measures detracking, not the whole reform. Ninth graders were affected too — they may have studied less once the exam mattered less. But because both groups experienced that during junior high, it cancels out in the difference-in-differences. So the paper identifies the net effect of the change in high school sorting, not the total effect of the reform package.
3. It cannot say why scores fell. It could be that more heterogeneous classrooms are harder to teach; it could be resource reallocation and peer effects. The paper cannot distinguish between them, or rule out that all of them contribute.