| Abstract: | 法國社會學家涂爾幹（1897）說「犯罪就是一種社會現象」，臺灣隨著社會變遷演變到M型化社會型態所造成的貧富不均問題，全球經濟景氣的聯動性，孕育許多犯罪種子，而不同犯罪類型主要成因皆不同。
本文將探討臺灣本島19個縣市於2007年1月至2022年1月經濟景氣、社會變遷因素與經濟犯罪（詐欺、違反著作權、查獲經濟案件）、暴力犯罪（強盜、強制性交）之相關性。經濟景氣以失業率、經濟成長率、景氣信號燈指數、歇業商店數等變數作為衡量指標，社會層面變數包括離婚率、人口密度、人口淨增加率、15歲以上高等教育程度、各縣市20-40歲人口比例、酒精銷售量等變數。
本研究以固定效果模型作實證分析，研究結果顯示失業率、經濟成長率、人口密度、人口淨增加率、各縣市20-40歲人口比率對於詐欺、違反著作權、查獲經濟案件犯罪率具有顯著影響；且經濟景氣因素對強盜犯罪率具顯著影響力，而人口密度、15歲以上高等教育程度同時對於強盜、強制性交犯罪具有顯著相關性。最後，再以近似不相關迴歸模型聯合分析各犯罪類型的相關性，結果顯示各犯罪行為的固定效果迴歸式之間的不可被觀察誤差項之間具有相關性。
;Durkheim（1897） said” Crime is a social phenomenon”. The National Policy Agency of Taiwan uncovered that the violent crime rate began to fall yearly; however, the economic crime rate has gradually increased over the past decades. Although the main causes of crime differ, most are related to social-economic factors. We are trying to discover the connection between social factors and the crime rate in Taiwan,
The research focused on Taiwan’s crime issues in 19 counties, from Jan. 2007 to Jan. 2022, based on Taiwan Public Statistics Database. The crimes include fraud, copyright infringement, economic crimes, robbery, and forced rape. Looking into the economic and social factors behind these crimes, we can find unemployment rate, economic growth rate, monitoring indicators, out of business, divorce rate, the density of population, net migration, high education, etc. The study applied the fixed-effects model and seemingly unrelated regression (SUR) model for analysis. Under the fixed-effect model, we found that the unemployment rate and economic growth rate, density of population, net migration, and the proportion of age 20-40 significantly impact economic crimes;.
In contrast, the unemployment rate, density of population, and high education factors significantly impact violent crimes. The results showed that both economic growth and recession would cause economic crimes. Nevertheless, the main causes of violent crimes are unemployment, social structure, and education. On the other hand, the SUR model shows that each type of crime is related. Therefore, the unobservable error terms of each regression are correlated. |
