| 摘要: | 本文主旨在探討失業率與實質薪資對於幸福指數之關係，觀察值為2012年至2015年，四年之資料，共觀測台灣20個縣市，以縱橫資料型態來做探究，並以兩種迴歸方法分別進行迴歸分析。
第一種方法為最小平方法，旨在觀察變數間的線性關係，本文得到以下結論：失業率和幸福指數之關係為負相關，所得對幸福指數之影響則是正相關，而另外添加的五項變數中，社會福利和受雇者與幸福指數之關係也是正向的，但歲入對幸福指數之影響則是呈現負向關係。將變數區分為南、北部後，可得知北部地區人民對於所得和幸福指數之關係，為顯著正相關，對於犯罪率和歲入之於幸福指數之關係，則是顯著負相關；南部地區人民對於總經變數之敏感程度較低，但對於社會變數都具顯著性。再將變數劃分為六都、非六都，可得到六都地區民眾教育程度與受雇者對幸福指數之關聯為顯著正相關，而歲入對幸福指數之關聯為顯著負相關；非六都地區民眾對總經指標較敏感，而社會變數方面，只有受雇者人數對幸福指數之關係具顯著性。
第二種方法為門檻迴歸法，旨在觀察變數間的非線性關係， 本文以社會治安、社會福利、教育程度、受雇者與歲入，五項社會指標作為門檻變數來探討失業率和所得個別對幸福指數之關係，結果是所有模型都不顯著。
The purpose of this paper is to explore the relationship between the unemployment rate and the real wage for the happiness index. The observation period is four years from 2012 to 2015. Survey 20 cities and counties in Taiwan, and use panel data to do the research with two different regression methods for analysis.
The first way is ordinary least square method, which aims to observe the linear relationship between variables: The relationship between the unemployment rate and the happiness index is negatively correlated, but the effect on salary and the happiness index is positively correlated. In the other five variables, the relationship between social welfare, employee percentage and the happiness index is positive, but the impact of the annual income on the happiness index is a negative relationship. Divide variables into North and South, we can see that for the relationship between income and happiness of the people in the North is positively correlated. However, relationship between crime rates and revenues to the happiness index is a significant negative correlation. People in the South is less sensitive to the Macroeconomic variable, but it is significant for the social variables.
Second method is threshold regression method, which aims to observe the non-linear relationship between variables. This paper discusses the relationship between the happiness index the unemployment rate and individual income by using social security, social welfare, education, employment and revenues, five social indicators, as threshold variables. Finally, the result is all models are not significant. |
