一篇論文的十年Ten years of one paper
一個問題，
我問了十年
One question,
ten years
2015 年 10 月，我為了申請研究所寫了一份研究計劃書，寫的是十二年國教免試入學。2025 年 10 月，同一個題目正式投稿到期刊——正好十年。中間它被改寫成三篇完全不同的文章，也被我放下很多次。
In October 2015 I wrote a research proposal to apply for graduate school. It was about Taiwan's high school admission reform. In October 2025 I submitted the same topic to a journal — exactly ten years later. In between, it was rewritten into three completely different papers, and put down many times.
起點：一個高中生留下來的疑問Where it started: a question left over from school
2015 年，十二年國教剛上路一年。當時吵得最兇的一件事是：如果真的「免試」入學，考試成績不再 100% 決定你上哪一所學校，那些傳統上「只收前面幾趴」的明星學校，還會是原來的樣子嗎？有人說門終於打開了，有人說學校只會用別的方式繼續挑人。
In 2015 the reform was one year old, and the loudest argument about it was this: if admission really became “exam-free” and test scores no longer decided 100% of where you went, would the schools that traditionally took only the top few percent still look the same? Some said the gate had finally opened. Others said schools would simply find another way to pick the same students.
我想知道答案。不是想知道「政策好不好」，而是想知道一件更具體、可以真的去量的事：一所高中裡面的學生，是不是還是被挑得一模一樣？如果一所學校只收前 3% 的人，那所學校裡的學生程度會非常整齊；如果門真的開了，那就會變得參差。所以「整齊或參差」這件事，本身就是答案。
I wanted to know. Not whether the policy was good or bad, but something more concrete and actually measurable: are the students inside a given high school still as alike as they used to be? If a school takes only the top 3%, the students inside it look very similar. If the gate really opened, the mix inside would get wider. So “alike or mixed” is itself the answer.
那時我手上有什麼？沒有資料庫，沒有統計訓練，什麼都沒有。所以計劃書裡我打算這樣做：
What did I have to work with? No database, no statistical training, nothing. So the proposal said I would do this:
- 發問卷。預計向 1,000 位高一學生發問卷，問他們當初怎麼填志願、最後上了哪裡。
- 比四個縣市。挑出基北、桃連、中投、高雄四個就學區，因為這四區在兩個年度之間「只有志願序的計分方式改變，其他項目都沒動」——我想用它們互相當對照。
- 自己造一把尺。「志願填得保守還是大膽」沒有現成的資料，所以我從招生簡章和教育部公布的序位百分比，硬推出一個指標。
- Run a survey. 1,000 first-year high school students, asking how they ranked their preferences and where they ended up.
- Compare four regions. Four school districts in which, between two admission years, only the preference-order scoring rule changed while everything else stayed put — so they could serve as each other's comparison.
- Build my own yardstick. No data existed on whether students ranked their choices cautiously or boldly, so I derived an index by hand from admission brochures and the Ministry's published percentile tables.
十年後我做同一件事，用的是完全不同的東西：六波 PISA 國際評量、四萬多名台灣學生，加上香港與澳門五萬多名學生當第二重對照。沒有問卷，沒有四個縣市，全部換掉了。
Ten years later I answered the same question with entirely different materials: six waves of PISA, more than forty thousand students in Taiwan, plus fifty thousand in Hong Kong and Macau as a second layer of comparison. No survey. No four districts. All of it replaced.
| 2015 年的我Me in 2015 | 2025 年的我Me in 2025 | |
|---|---|---|
| 想知道的事What I wanted to know | 一所學校裡的學生，是不是還是被挑得一模一樣Whether the students inside a school are still picked to be alike | 一所學校裡的學生，是不是還是被挑得一模一樣——如果不是的話，誰受益？誰受害？Whether the students inside a school are still picked to be alike — and if not, who gains and who loses? |
| 資料哪裡來Where the data came from | 打算自己發 1,000 份問卷；蒐集招生簡章與序位百分比表進行分析A survey of 1,000 students I would run myself, plus admission brochures and percentile tables collected and analysed by hand | 六波 PISA 國際評量，41,465 名台灣學生Six waves of PISA, 41,465 students in Taiwan |
| 跟誰比Compared against whom | 不同就學區互相比，因為它們只差一件事（這在比較政治領域叫「最大相似體系比較」，most similar systems design, MSSD）Different school districts against each other, because only one thing differed between them (comparative politics calls this a most similar systems design, MSSD) | 高中生跟國中生比，再跟港澳比（這在計量經濟學叫 DID 與 DDD）High school against junior high, then against Hong Kong and Macau (econometrics calls this DID and DDD) |
最有意思的是後面那兩列，資料來源與分析工具全部換過。可是「想要了解這個政策到底改變了什麼」這個核心問題沒有變。研究問題還擴張了：從單純想知道「一所學校裡的學生，是不是還是被挑得一模一樣」，到去問一個政策有沒有達成它自己訂下的目標，以及會付出什麼代價。
The interesting rows are the last two: the data sources and the analytical tools were replaced wholesale. What did not change is the core question — what did this policy actually alter? If anything the question grew: from simply asking whether the students inside a school are still picked to be alike, to asking whether a policy met the goal it set for itself, and what it cost to get there.
成熟的是技術，
與問問題的精準度。
不變的是初衷。
The technique grew up,
and so did the precision of the asking.
The question never changed.
中間發生的事What happened in between
一份為了申請研究所寫的計劃書
A proposal written to apply for graduate school
問卷、四個就學區、一把自己造的尺。那時候我還不知道自己想做的事有名字。
A survey, four school districts, and a yardstick I built myself. I did not yet know that what I wanted to do had a name.
然後我沒有考上那一所研究所。這份計劃書沒有完成它被寫出來的任務。
And then I did not get into that programme. The proposal failed at the job it was written for.
在立法院工作
Working at the Legislative Yuan
兩年沒有碰這個題目——但沒有離開這個領域。我在立法院用另一種方式關心教育議題：不是寫研究，是看預算、看法規、看一個政策在真的執行時會卡在哪裡。
Two years without touching this topic — but never away from the field. At the Legislative Yuan I worked on education from a different angle: not writing research, but reading budgets and regulations, and watching where a policy actually gets stuck once it has to be implemented.
政大公行系碩士班
Master's programme, public administration
兩年後我進了碩士班——不是當初申請的那一所，寫的也不是那份計畫。這兩年我做的是別的教育研究：教育對階級流動的影響、大學教學支出對學生就業與薪資的影響。那個初衷被我擱在一邊。
Two years later I did get into a master's programme — not the one I had applied to, and not with that proposal. What I worked on there was other education research: how education shapes class mobility, how university instructional spending affects graduates' employment and earnings. The original question stayed on the shelf.
但這兩年一點都沒有白費。我在這裡學會怎麼做研究、怎麼處理次級資料、什麼叫因果推論、什麼是雙重差分。八年後真正把那個題目做出來所需要的工具，是在這兩年拿到的。
Those two years were not wasted. This is where I learned how to do research at all — how to work with secondary data, what causal inference means, what difference-in-differences is. The tools that eventually made the old question answerable were picked up here.
政策評估課的期末報告：先問「隔離」
A Program Evaluation term paper: asking about segregation first
進入博士班之後，我第一次真的把這個題目做出來，是在一堂政策評估（Program Evaluation）課上。但我問的不是後來那個問題，而是：改革有沒有讓「成績相似的學生集中在同一所學校」這件事變少？那份報告的標題直接叫〈打破明星學校的迷思？〉。
Once I started my PhD, the first time I actually carried this out was for a Program Evaluation course. But the question was not the one I ended up publishing. It was: did the reform reduce the extent to which similar students end up concentrated in the same schools? The paper was titled “Break the myth of school ranking?”
我用了社會學衡量「隔離」的標準工具（Duncan 相異指數），同時看兩件事：學生按成績被分開的程度，以及按家庭背景被分開的程度。結論是能力隔離下降約 20%，而且都會地區的效果是其他地區的兩倍。
I used the standard sociological tool for measuring segregation (the Duncan dissimilarity index), and looked at two things at once: how far students were separated by achievement, and how far by family background. The reform reduced ability segregation by around 20%, with the effect in urban areas about twice that elsewhere.
然後我停了下來。這篇沒有再往前走。
And then I stopped. That paper never went any further.
換一堂課，換一個問法：改問「分流」
Another course, another framing: asking about tracking instead
一年後的另一份課堂報告（Advanced Empirical Methods），我把問題換了個方向。「隔離」問的是學生有沒有被分開；「分流」問的是把不同程度的學生放在一起，對他們的學習到底是好是壞。這一問，就從社會學的地盤走進了教育經濟學吵了幾十年的那場架。
A year later, in another term paper (Advanced Empirical Methods), I turned the question around. “Segregation” asks whether students are separated. “Tracking” asks whether putting students of different levels together is good or bad for their learning. That one shift moved the paper out of sociology and into a decades-old argument in the economics of education.
這一版第一次寫出了後來那篇論文的形狀。它的一則註腳寫著：「我先前在另一份學期報告的分析，沒有發現改革會加劇或消除校際的社經隔離。」那個註腳指的就是支線一——而它一路活到了刊出版，還從一則註腳升格成正文表格裡的一欄。被放棄的那篇沒有白寫，它只是換了一個身分留下來。
This was the first version with the shape of the published article. One of its footnotes reads: “my previous analysis (in another term paper) provides no evidence that the reform would aggravate or eliminate the socio-economic segregation across schools.” That footnote is pointing at branch one — and it survived all the way into the published article, promoted from a footnote to a column in a table in the main text. The abandoned paper was not wasted. It just carried on under a different identity.
同一批資料，改寫成一篇公共行政論文
Same data, rewritten as a public administration paper
投到台灣公共行政與公共事務系所聯合會（TASPAA）的研討會，題目叫〈上有政策，下有對策？〉。用的是同一份 PISA、同一套比較邏輯，但問了一個經濟學設計問不出來的問題：學校會不會用不違法的方式閃避政策？
Presented at the TASPAA conference under the title “Policies from above, countermeasures from below?” Same PISA data, same comparison logic — but asking something the economics version could not: would schools work around the policy without ever breaking a rule?
結論是會。學校並沒有違反教育部的規定，卻透過其他法規允許、但教育部未必樂見的做法，偏離了政策的方向。換一個學科去問同一批資料，會問出完全不同的問題。
They would. Schools broke no Ministry rule, yet drifted away from the policy's direction using routes other regulations permitted and the Ministry would not have welcomed. Ask the same data from a different discipline and a completely different question comes out.
「你在我課上寫的那篇後來怎麼了？」
“What ever happened to the paper you wrote for my class?”
我把博士論文的研究計畫寄給口試委員。其中一位回信，問了一個和計畫完全無關的問題。那封信在下面。
I sent my dissertation proposal to my committee. One of them replied with a question that had nothing to do with the proposal. That email is below.
寫博士論文、找工作
Dissertation, then the job market
博士論文寫的是另外三個題目，接著是求職。這個檔案就停在那裡，三年多沒有真正動過。
The dissertation was about three other topics, and then came the job market. This file simply sat there for more than three years, essentially untouched.
已經是助理教授的我，把它打開
By then an assistant professor, I opened the file
重寫，然後在 10 月投出去。從那份研究計劃書算起，正好十年。
I rewrote it, and submitted in October. Counting from that first proposal: exactly ten years.
兩位審查人、兩輪修改
Two referees, two rounds of revision
《人文及社會科學集刊》接受，並於 2026 年 6 月正式在期刊網站上刊登。
Accepted by the Journal of Social Sciences and Philosophy, and published on the journal's site in June 2026.
六個版本，是六篇不同的文章Six versions, six different papers
「同一個題目寫了十年」聽起來像是同一篇文章改了很多次。其實不是。我把其中五個版本兩兩比對，計算它們有多少連續八個字完全相同的片語——這個指標很嚴格，只要換一個詞就不算。（2015 年那份計劃書是中文寫的，而且從來沒有真的執行，沒有可以比對的文字，所以不在這張圖裡，但列在下面的表中。）
“Working on one topic for ten years” sounds like revising one paper over and over. It was not. I compared five of the versions pairwise, counting strings of eight consecutive identical words — a strict measure, since changing a single word breaks the match. (The 2015 proposal was written in Chinese and was never actually carried out, so there is no comparable text; it is absent from the chart but present in the table below.)
五個版本之間，有多少文字是共用的？How much text do the five versions actually share?
顏色越深＝重疊越多。數值為兩篇共有的八字片語，佔較短那篇的比例Darker = more overlap. Values are shared 8-word strings as a share of the shorter paper
| 版本Version | 標題Title | 核心問題Core question | 方法工具Tools | 主要結論Main conclusion |
|---|---|---|---|---|
| 2015-10 計劃書2015-10 Proposal 研究所入學申請grad school application |
十二年國教免試入學超額比序志願序計分方式之影響The effect of preference-order scoring in the exam-free admission system | 改革有沒有讓一所學校的學生不再一模一樣？Did the reform stop schools from being filled with students who are all alike? | 自發問卷、政府官方政策文件與統計蒐集、跨縣市比較A self-run survey, official policy documents and statistics, cross-district comparison | ——（從未執行）— (never carried out) |
| 2020-11 隔離報告2020-11 Segregation 政策評估課course paper |
打破明星學校的迷思？入學改革對能力與家庭背景隔離的影響Break the myth of school ranking? The impact of admission reform on school segregation by ability and family background | 改革有沒有讓成績相似的學生比較不集中在同一所學校？Did the reform reduce the concentration of similar students in the same schools? | Duncan 相異指數、雙重差分、bootstrap 標準誤、兩類安慰劑檢定Duncan dissimilarity index, DiD, bootstrapped SEs, two placebo tests | 能力隔離下降約 20%；都會地區效果是其他地區的兩倍Ability segregation fell about 20%; the urban effect was roughly twice that elsewhere |
| 2021-11 分流報告2021-11 Tracking 課堂報告course paper |
分流對學生表現水準與分布的影響——來自台灣入學改革的證據The Effect of Tracking on Student Performance Level and Distribution — Evidence from Admission Reform in Taiwan | 把不同程度的學生放在一起，對學習是好是壞？誰受益、誰受損？Is mixing achievement levels good or bad for learning — and who gains, who loses? | 雙重差分、事件研究、分量迴歸；改用校內成績落差當分流的量尺DiD, event study, quantile regression; switched the yardstick to within-school spread | 校內落差擴大、平均表現下降，而且整個分布一起降Within-school spread widened, average performance fell, and the whole distribution fell together |
| 2022-05 研討會2022-05 Conference TASPAA・公共行政TASPAA, public admin |
上有政策，下有對策？政策順服對政策效果之影響Policies from above, countermeasures from below? Policy compliance and policy effectiveness | 學校會不會用不違法的方式閃避政策？什麼樣的學校比較會閃？Do schools work around the policy without breaking rules — and which schools do? | 同一套雙重差分，但依變項換成學校自己的行為（入學是否採計在校成績、是否校內能力分組）The same DiD, but the outcome becomes the school's own behaviour (using school grades in admission; running ability grouping) | 學校沒有違規，卻用其他法規允許的做法偏離政策方向；這種「陽奉陰違」確實損害政策成效Schools broke no rule yet drifted from the policy using other permitted routes; this covert non-compliance did damage the policy's effect |
| 2025-10 投稿版2025-10 Submitted |
公平與效率不可兼得？免試入學對學校分層、教育資源分配與學生表現之影響The Equity–Efficiency Trade-off: Impact of Taiwan's High School Detracking Reform… | 公平的改善，是不是以效率為代價？Did the equity gain come at the cost of efficiency? | 再加上三重差分（港澳）、資源可及性分析Adds a triple difference (Hong Kong/Macau) and a resource-access analysis | 分層下降、後段學生資源改善，但平均表現下降 0.171 個標準差Stratification fell and low achievers gained resources, but average performance fell 0.171 SD |
| 2026-06 刊出版2026-06 Published 人文及社會科學集刊J. Soc. Sci. & Philosophy |
十二年國教免試入學的政策效果：教育公平改革是否以犧牲學生學習成就為代價？Does Reducing Ability Tracking Improve Equity at the Cost of Student Achievement? | 公平的改善，是不是以學生成就為代價？（同一個問題，換掉一個詞）Did the equity gain come at the cost of student achievement? (same question, one word changed) | 同上，再加 PISA 合理值處理與 BRR 標準誤、熵平衡重加權As above, plus PISA plausible values, BRR standard errors and entropy balancing | 校內落差 +30 分（補回約一半的分流差距）、平均總分 −50 分（−0.18 SD），整個分布一起降Within-school spread +30 points (about half the tracking gap), average total −50 points (−0.18 SD), uniform across the distribution |
六行的「標題」「方法」「結論」全都不一樣，只有一件事沒變：每一行都在問「這個制度改變之後，誰跟誰被放在一起，結果怎麼樣」。換過三個學科的問法、四套工具、兩種語言，那個核心問題一次都沒被換掉。
Across six rows, the titles, the tools and the conclusions all differ. Only one thing holds: every row asks who ends up with whom after the rules change, and what happens as a result. Three disciplinary framings, four toolkits, two languages — and the core question was never replaced.
那封信The email
2023 年 1 月 30 日，我把博士論文的研究計畫寄給五位口試委員。兩天後，其中一位回信。信的開頭是禮貌性的一句「Thanks for this」，然後就轉了個彎：
On 30 January 2023 I emailed my dissertation proposal to my five committee members. Two days later, one of them replied. It opened with a polite “Thanks for this,” and then turned:
日期：2023 年 2 月 1 日
主旨：RE: Dissertation Proposal Defense — Yung-Yu Tsai From: a committee member (an economist of education)
Date: 1 February 2023
Subject: RE: Dissertation Proposal Defense — Yung-Yu Tsai
Yung-Yu,
Thanks for this. A quick question—what ever happened to the paper you wrote for my class? I am struggling to remember the precise topic (peer effects, grade inflation?), but my recollection is that it was very good 🙂
他甚至不太記得那篇在寫什麼——他猜了「同儕效應？成績膨脹？」，兩個都不對。他記得的不是題目，是那篇東西好。兩個星期後的口試現場，他又問了我一次。
He could not quite remember what it was about — he guessed “peer effects, grade inflation?”, and both were wrong. What he remembered was not the topic. It was that the paper was good. Two weeks later, at the proposal defence, he asked me again.
說實話，當時我心裡覺得那篇不夠好。它只是一份學期報告，識別策略沒站穩，穩健性檢查沒做完，我自己知道有幾個洞。所以我沒把它放進博士論文。但「有一個人喜歡這篇文章」這件事，我記了兩年多。
Honestly, at the time I thought the paper was not good enough. It was a term paper; the identification was shaky, the robustness checks were unfinished, and I knew where the holes were. So I left it out of the dissertation. But the fact that somebody liked it stayed with me for more than two years.
這封信只有兩行，寫的人大概三十秒就寄出去了，而且連題目都記錯。但它是這篇論文最後能被寫完的原因之一。你隨口說的那句「這篇寫得很好」，對方可能會記很多年。
This email is two lines long. Whoever wrote it probably spent thirty seconds on it, and got the topic wrong. It is also one of the reasons this paper eventually got finished. The offhand “that was very good” you say to a student may be carried around for years.
十年裡，我改了什麼，守住了什麼Over ten years: what I changed, what I held
一篇文章寫十年，會經過很多人的手。這十年裡，至少有四批人碰過它：兩堂課的老師、研討會上的評論人、其他請益諮詢過的老師，以及最後兩位匿名審查人與責任編輯。每一批人都要我改東西，而且他們要的東西不一樣。
A paper that takes ten years passes through a lot of hands. At least four sets of people touched this one: the instructors of two courses, the discussant at the conference, other teachers I went to for advice, and finally two anonymous referees and the handling editor. Each group wanted something changed, and each wanted something different.
回頭看，被改掉的幾乎都是技術與名字；沒有被改掉的，是我一開始想知道的那件事。
Looking back, what got changed was almost entirely technique and vocabulary. What did not get changed was the thing I wanted to know in the first place.
我改掉的What I changed
- 資料。從打算自己發 1,000 份問卷，換成六波 PISA、四萬多名學生。政府公告的招生簡章沒有用上一張。
- 量尺。從社會學的隔離指數，換成教育經濟學的校內成績落差。同一件事，換一個學科的量法。
- 方法。從「挑幾個只差一件事的縣市互相比」，換成雙重差分、三重差分、事件研究、分量迴歸。誤差的算法也從沒想過，換成 bootstrap，再換成 BRR 加合理值處理。
- 學科身分。社會學問法 → 經濟學問法 → 公共行政問法 → 再回到經濟學問法。公行那一次不是繞路，它讓我看見一件經濟學設計看不見的事：學校會閃。
- 概念的名字，兩次。2021 年把「隔離」改成「分流」；2026 年把「效率」改成「學生成就」。兩次都不是文字修飾，是承認自己原本量的東西配不上原本用的那個詞。
- The data. From a survey of 1,000 students I would run myself, to six waves of PISA and forty thousand students. Not one of the government admission brochures I collected was ever used.
- The yardstick. From a sociological segregation index to the economics-of-education measure of within-school spread. Same phenomenon, a different discipline's ruler.
- The method. From “pick a few districts that differ in one thing and compare them” to difference-in-differences, triple difference, event study and quantile regression. Standard errors went from an afterthought, to bootstrapping, to BRR with plausible values.
- The disciplinary identity. Sociology → economics → public administration → back to economics. The public administration step was not a detour: it showed me something an economics design cannot see, namely that schools dodge.
- The name of the concept, twice. “Segregation” became “tracking” in 2021; “efficiency” became “student achievement” in 2026. Neither was cosmetic. Both were admissions that what I was measuring did not deserve the word I had used for it.
我守住的What I held
- 要量的那件事。一所學校裡的學生，是不是被挑得一模一樣。2015 年的計劃書叫它「入學成績之分布情形」，2020 年的課堂報告叫它隔離指數（segregation index），2026 年的論文叫它 within-school standard deviation。同一個東西，換了三次名字。
- 比較的邏輯。找兩群只差一件事的人。從四個就學區，到國中對高中，再到台灣對港澳——形式一直換，念頭沒換過。
- 誰是主角。那些原本進不去、後來進得去的孩子。每一個版本都在問他們身上發生了什麼。
- 真正的初衷。想知道一個立意良善、目標遠大的政策，是不是真的達成了它宣稱的事，以及可能帶來哪些負面影響——這個原初的動機，十年沒有變過。
- The thing being measured. Whether the students inside a school are picked to be alike. The 2015 proposal called it “the distribution of admitted students' scores”; the 2020 term paper called it a segregation index; the 2026 paper calls it the within-school standard deviation. Same quantity, renamed three times.
- The comparison logic. Find two groups that differ in one thing. Four districts, then junior high against high school, then Taiwan against Hong Kong and Macau — the form kept changing, the instinct never did.
- Who the paper is about. The students who would not have got in before and do now. Every version asks what happened to them.
- The original impulse. Wanting to know whether a well-intentioned policy with an ambitious goal actually achieved what it claimed, and what harms it might carry with it. That motive has not changed in ten years.
十年裡我換掉了資料、量尺、方法、學科，甚至換掉了描述這件事的詞。能換的都換了。剩下沒換的那個東西，就是初衷。你不會在寫作的當下認出它——只有在回頭看的時候，才會發現有一樣東西從頭到尾沒被動過。
Over ten years I replaced the data, the yardstick, the method, the discipline, even the words used to describe the thing. Everything replaceable was replaced. What was left over is the original impulse. You cannot recognise it while you are writing — only looking back do you notice the one thing nobody ever touched.
如果你正在讀書If you are still a student
我把這個過程寫下來，是因為學術訓練很少告訴你這件事：研究可以走得非常長，而且中間可以停很久。
I wrote this down because academic training rarely tells you this: research can take a very long time, and it is allowed to stop for most of it.
- 年少時代的一點悸動，是可以當真的。2015 年我是為了申請學校才寫那份計劃書（而且還沒申請上）。它不是什麼偉大的構想，就是一個高中生活留給我的、沒被回答的疑問。十年後，它是一篇期刊論文。
- 放下不等於放棄。這個題目我放下過很多次：從研究所申請計劃書到真的寫成課堂報告，隔了五年；從研討會發表到正式投稿期刊，又等了三年半。它就在那裡等，沒有壞掉。
- 寫不完的東西不是浪費。2020 年那篇「隔離」報告，最後只變成另一篇文章裡的一則註腳，以及刊出版正文表格裡的一欄。但如果沒有它，我不會知道該把問題轉向「分流」。
- 同一個問題可以有很多種身分。同一批資料，我寫過社會學版、經濟學版、公共行政版；同一個現象，我當過「隔離」也當過「分流」。換一個學科去問它，你會問出不同的問題。
- 回去的時候，你帶著新的工具。這才是重點。不是「我終於有時間寫完它了」，而是「我現在終於配得上這個問題了」。2015 年的我做不出這篇論文——不是因為不夠努力，也不是因為我問錯問題，是因為那時我還沒有那組工具。
- A small stirring from your school years can be taken seriously. I wrote that 2015 proposal to get into a programme (and it did not get me in). It was not a grand idea — just an unanswered question left over from my own time at school. Ten years later it is a journal article.
- Putting something down is not giving it up. I put this topic down many times: five years passed between the application proposal and the first real term paper, and another three and a half between the conference presentation and the journal submission. It waited. It did not spoil.
- Unfinished work is not wasted work. The 2020 segregation paper ended up as a single footnote inside a different article — and as one column of a table in the published version. But without it I would never have known to turn the question toward tracking.
- One question can hold several identities. With the same data I wrote a sociology version, an economics version and a public administration version; the same phenomenon was “segregation” once and “tracking” another time. Ask it from a different discipline and you will get a different question out of it.
- When you go back, you go back with better tools. That is the real point. Not “I finally had time to finish it,” but “I am finally adequate to it.” The 2015 version of me could not have written this paper — not for lack of effort, and not because the question was wrong, but because I did not yet have the tools.
那個問題一直在原地。
移動的是我。
The question never moved.
I did.