罰單與道安站標FINESROADSAFETY

罰單變成了誰的收入Where fine revenue goes

這一站只回答一個問題:地方政府收了罰單的錢之後,有多少變成道路工程,有多少留在帳上。每一格都標出處,查不到就寫 pending。One question only: of the fine revenue a local government collects, how much turns into road engineering and how much stays on the books. Every cell is sourced; unknown is written as pending.

114 年度
205 億
分配給地方政府的罰鍰收入fine revenue allotted to local governments
草案
12% → 100%
法定提撥率(第 4 條)statutory earmark ratio (Art. 4)
淨增加
180 億
205 × 88%,不是總額205 × 88%, not the total
上路
2028pending
發布施行次一會計年度;截至 11/03 尚未發布next fiscal year after promulgation; not promulgated as of 11/03

為什麼是這一題Why this question

2026 年 10 月交通部預告把地方罰鍰的提撥率從 12% 拉到 100%,官方理由是撕掉「罰單當提款機」的標籤。但如果只看這條新聞,會得到一個錯的印象:好像誘因被切斷了。實際上被改的是用途,不是來源。In October 2026 MOTC proposed raising the local earmark from 12% to 100%, framed as ending the use of fines as a cash machine. Read only the news and you get the wrong impression: the incentive is intact; only the label on the money changed.

所以要量的是三個方向有沒有背離:開單的量、收到錢的量、變成工程的量。三者同向=罰單確實被當收入;反向=執法有效。So we measure whether three directions diverge: tickets issued, money collected, engineering delivered. Same direction means fines really are revenue; opposite means enforcement works.

機制:分錢的條文與用錢的條文是兩條Mechanism: two different articles

「罰單當收入」的引擎在第 3 條;這次修的是第 4 條。前者決定錢往誰那裡去,後者只決定錢被貼上什麼用途標籤。The engine behind fine-as-revenue is Article 3. The amendment touches Article 4. The first decides who gets the money; the second only labels what it may be spent on.

條文article內容content這次有沒有動touched?
第 3 條 地方舉發案件:75% 分配予直轄市/縣(市)政府、24% 分配予各處罰機關、1% 解繳國庫Locally-issued cases: 75% to the city/county, 24% to the sanctioning agency, 1% to the treasury 未動
第 4 條 現行:應至少提撥 12%,六款「優先支應」項目(舉發處理、裝備器材、資訊電腦化、保管場、宣導、教育訓練)Current: at least 12%, six priority items (case handling, equipment, IT, impound lots, publicity, training) 修正中
草案第 4 條 提撥率 100%;新增「道路交通安全環境改善費用」與「交通事故處理與分析費用」;每半年公開編列與執行100% earmark; adds road-environment improvement and crash analysis items; semi-annual public reporting 預告 2026-10-02
未刪除的一項what stays 科技執法設備仍屬支應項目——交通部明說「未刪除設置科技執法設備用途」Tech enforcement equipment remains eligible: MOTC stated the item was not removed 保留

結論先寫在這裡:100% 專款專用改變的是用途,不是誘因。要切斷「多開單=多收入」,得動的是第 3 條的 75%,或把績效指標從取締件數換掉。Stated up front: 100% earmarking changes the use of money, not the incentive. Cutting 'more tickets = more revenue' requires changing the 75% rule in Article 3, or replacing ticket counts as the performance metric.

四個指標,回答「搞事」而不是「違法」Four metrics that answer 'gaming', not 'illegality'

罰鍰當收入不違法——它是預算制度的一部分。所以問題不是合不合法,是四個比率有沒有背離。每一格都要能回查原始資料。Treating fines as revenue is not illegal; it is part of the budget system. The question is whether four ratios diverge. Each cell must trace back to raw data.

編號id指標metric算法formula資料源source狀態state
M1取締強度Enforcement intensity舉發件數 ÷ 人口警政署 舉發統計(data.gov.tw 39362)ready
M2收入對支出比Revenue vs works罰鍰分配金額 ÷ 該縣市道安工程決標金額公路局分配金額 CSV + 標案 APIready
M3方向背離Direction split罰單年增率 vs 死傷年增率警政署 + 道安網 368 鄉鎮ready
M4自償循環Self-funding loop科技執法設備決標金額 vs 隔年罰單成長標案 API + 舉發統計ready

M2 是這一站的主指標:收入有官方 CSV、支出有逐案決標,兩邊一 join 就是「收了多少、花在哪」。它不需要等 2028 的半年報。M2 is the headline: revenue has an official CSV, spending has case-level awards. Joining them answers 'how much collected, where spent' without waiting for 2028.

大桃園:13 區的道安工程實際發到哪裡Taoyuan: where the road-safety works actually went

這是「支出」這一側的完整樣本:用 16 組關鍵字反向撈桃園市 2015–2026 的決標公告,逐案歸到 13 區或市府局處。每一筆都有決標日、機關、金額、案號。The spending side, end to end: 16 keywords harvest Taoyuan awards from 2015 to 2026, attributed case by case to one of the 13 districts or to a city bureau. Every row has date, agency, amount and case number.

件數
667
2015–2026 決標awards 2015–2026
金額合計
68.3 億
含開口契約上限,非實支includes open-contract ceilings
市府局處
56.1 億
跨區工程歸這裡cross-district works
13 區小計
12.2 億
區公所自行發包awarded by district offices

13 區 + 市府局處13 districts + city bureaus

歸屬unit件數cases金額(萬元)NT$1k最近決標latest
桃園區3629,99220260108
平鎮區4516,49820251219
蘆竹區2413,40920260810
八德區2711,27020261020
龜山區389,62020260723
中壢區257,91420260128
龍潭區286,61220260924
大園區196,58320260716
楊梅區235,72320261014
大溪區115,19320250627
觀音區274,36220260123
新屋區243,94920260901
復興區51,34720251224
市府局處335560,58820261021

逐年by year

年year件數cases金額(億)NT$100M平均每件(萬)avg/case
2026416.561,600
2025597.901,339
2024716.38898
2023608.711,451
2022615.97979
2021574.21738
2020646.871,073
2019443.25739
2018575.911,036
2017515.481,074
2016493.95806
2015433.14729

金額最大的 8 件largest 8 awards

  • 20250825 市府局處 桃園市桃園區成功二橋新建工程暨成功路五叉路口改善工程 30,221 萬
  • 20260303 市府局處 115年度桃園市人行道及通學廊道改善與修繕工程(開口契約) 24,975 萬
  • 20231229 市府局處 113年度桃園市人行道改善及修繕工程(開口契約) 20,000 萬
  • 20180124 市府局處 107年度人行道巡查及修繕工程開口契約 15,281 萬
  • 20220113 市府局處 桃園市111年度交通標誌標線工程(開口合約) 12,660 萬
  • 20170104 市府局處 106年度人行道巡查及修繕工程開口契約 10,380 萬
  • 20171227 市府局處 桃園市107年度交通號誌工程(開口合約) 10,183 萬
  • 20240123 市府局處 桃園市重點場域暨高鐵聯絡道周邊路網 AI號誌控制示範計畫第三期 10,000 萬

資料抓取於 2026-11-03 15:16。三件事先講清楚:①開口契約的決標金額是上限,實際支出要等結算;②這是關鍵字樣本,不是桃園全部工程預算;③罰鍰收入只到市層級,13 區沒有各自的罰單收入,所以這一頁只呈現「花了多少」,不呈現「收了多少」。Fetched 2026-11-03 15:16. Three caveats: open-contract amounts are ceilings; this is a keyword sample, not the whole city budget; fine revenue exists only at city level, so this page shows spending, not collection.

拿來對照的那個數字:桃園 114 年度獲配罰鍰約 25.1 億(交通部口述)——是同年道安工程樣本 6.6 億的近 4 倍。但同一份官方 CSV 記的是 1.56 億,兩者差 16 倍,口徑未定前不要下結論。For reference: Taoyuan received about NT$2.51bn in fines in FY2025 (MOTC), nearly 4x the NT$0.66bn works sample the same year. The official CSV instead records NT$0.156bn — a 16x gap. Do not conclude until the basis is settled.

建置進度Build progress

3/12 完成。這張表由 data/units.json 產生,不是手寫;每一項有自己的驗收方式。3/12 done. Generated from data/units.json, not typed; each unit has its own check.

單元unit內容scope狀態state備註note
U1開站骨架與部署shell + deploydone2026-11-03 v0.1 骨架,四道閘門(version.json/版本戳/seo_audit/首頁<400KB)
U2機制章(第3條 vs 第4條)mechanism chapterdone法規資料庫現行版全文 + 公報預告文,兩處都已引用
U3指標定義與資料源metric definitionsdoneM1–M4 四指標,每個都有公式與來源;資料源已實測可抓
U4來源存檔(sha256)source registrydoing第一批:公報預告頁、法規資料庫第3/4條、交通部新聞稿、公路局分配金額 CSV
U5罰鍰分配金額 ETLrevenue ETLtodo公路局開放資料 CSV,22 縣市 × 年
U6舉發件數 ETLticket ETLtodo警政署 舉發統計(data.gov.tw 39362),已下載驗證 ODS 格式
U7道安決標 ETLawards ETLtodotwbuying/openfun 免金鑰 API,20 組關鍵字反向撈
U8死傷 ETL(368 鄉鎮)casualty ETLtodo道安資訊查詢網,鄉鎮市區層級、月更到 2 個月前
U9四比率 join 與排序ratio jointodoM1–M4 逐縣市排序,每格標資料源與抓取日
U10368 鄉鎮檢視township viewtodo錢只到 22 縣市;鄉鎮層用標案金額拼
U112028 半年報解析器semi-annual parsertodo22 縣市 22 種格式,正規化欄位;2028 後才會有資料
U12每週自評與監看weekly audittodoseo_audit + 來源變動偵測,掛 cron

來源清冊Source registry

0 筆已存檔。機器可讀清冊在 /archive.json。0 archived. Machine-readable index at /archive.json.

v0.1 尚無存檔來源。第一批來源在 U4 落地後出現在這裡,每一筆會有 URL、抓取時間與 sha256。No archived sources in v0.1. The first batch appears here after U4, each with URL, fetch time and sha256.

怎麼量:每格都要能回到原始資料How we measure: every cell traces back

  • 收入Revenue
    公路局開放資料「道路交通違規罰鍰收入各縣市分配金額」CSV,22 縣市 × 年度。Highway bureau open-data CSV: fine revenue allotted per county by year.
  • 開單量Tickets
    警政署「舉發違反道路交通管理事件成果統計表」(data.gov.tw 39362),含年度、違規態樣、法條、各縣市件數。NPA enforcement statistics (data.gov.tw 39362) by year, violation type, statute and county.
  • 支出Spending
    政府採購決標公告逐案(免金鑰第三方 API),以標線/號誌/人行道/測速/科技執法等約 20 組關鍵字反向撈,再按機關與縣市歸戶。Case-level awards via a keyless third-party API, harvested by ~20 keywords then attributed to agency and county.
  • 結果Outcome
    道安資訊查詢網的鄉鎮市區死傷統計,368 個鄉鎮、每月更新到兩個月前。Road-safety portal casualty counts for all 368 townships, monthly with a two-month lag.
  • 驗收Check
    每個比率都由第二支獨立查詢重算一次,對不上就查欄位語意,不上色。Every ratio is recomputed by a second independent query; mismatches are investigated, not painted over.

這一版還缺什麼What this version lacks

  • 草案不等於定案Draft is not law
    第 4 條修正仍在預告階段(意見截止 2026-10-16),條文與比例都可能變動;引用時請寫「草案」。Art. 4 is still a draft (comments closed 2026-10-16). Cite it as a draft.
  • 半年報不存在No semi-annual reports yet
    2028 之前沒有任何縣市會公布「罰鍰收入執行進度」,這一格現在無法量,只能量收入與決標。Before 2028 no county publishes earmark execution; only revenue and awards are measurable now.
  • 標案資料是第三方Awards data is third-party
    目前用 twbuying/openfun 兩個免金鑰 API 取得決標公告,兩者都是爬政府電子採購網;官方沒有逐案 API。正式引用要回採購網原頁。Awards come from two keyless third-party APIs that crawl the official site; there is no official case-level API.
  • 鄉鎮市區沒有罰鍰收入Townships have no fine revenue
    罰鍰只分配給 22 個縣市,368 個鄉鎮市區不是受配單位;鄉鎮層級的「花了多少」只能用標案金額拼出來。Fines go to the 22 counties only; the 368 townships receive none. Township spending can only be rebuilt from awards.

引用規則How to cite

可以引用站上的數字,但要連回出處;標 pending 的格子不是事實,請勿當結論轉述。You may cite the numbers, with the source link. Cells marked pending are not facts.

版本 2026.11.03-89155a8 · build d665efdb55ee · 建置 2026-11-03 15:16 UTC+8 · /version.json

關於這個站About

一個人維護,跟 sr71.sean-hsu.com 同一條生產線:靜態生成、來源存檔存 sha256、每週自評、內容雜湊版本。不作政策倡議,只把可回查的比率放上桌。One person, same pipeline as sr71.sean-hsu.com: static build, sha256 source registry, weekly self-audit, content-hash versions. No advocacy, just checkable ratios.