犯罪數與宣告刑總和對執行刑訂定的影響

Abstract

本文以2013至2023年臺灣法院裁判書資料，對數罪併罰案件定執行刑模式進行實證分析，探討「犯罪數」與「各罪宣告刑期加總」對「執行刑占法定上限之比例（執行刑占比）」及「執行刑長度」的影響。研究透過Python蒐集與整理六組資料集，以普通最小平方法（OLS）與固定效果模型（FE）估計，並以斷點迴歸等方法評估2018 年頒布的「刑事案件量刑及定執行刑參考要點」，及 2023 年「評價型與事實型量刑資訊系統」啟用之政策效果。結果顯示，多數樣本中犯罪數與宣告刑期加總均與執行刑占比呈顯著負相關，但與執行刑長度呈顯著正相關；顯示當犯罪數較少或各罪宣告刑期較輕時，法官較傾向裁定接近各罪宣告刑單純加總之執行刑(即執行刑占比相對較高)，而犯罪程度較嚴重者會得到更長的執行刑。至政策檢驗則未呈現一致顯著的斷點或變異減少效果，顯示政策對定執行刑決策之可辨識影響有限。

Using Taiwanese court judgments from 2013–2023, this study examines sentencing patterns in multiple-offense cases, focusing on how the number of offenses and the sum of declared sentences affect (i) the ratio of the executed sentence to the statutory upper bound and (ii) the executed-sentence length. Based on six datasets compiled with Python and estimated with OLS and fixed effects, the results show a robust negative association between the key predictors and the ratio, but a positive association with executed-sentence length. Judges appear to set relatively higher ratios in cases with fewer offenses or shorter total sentences, consistent with anchoring on the upper bound, while more severe cases receive longer executed sentences. Tests around the 2018 and 2023 reforms reveal no consistent discontinuities or reductions in dispersion, suggesting limited identifiable policy effects.