2010年美國居高不下的失業率使得身為執政黨的民主黨在國會期中選舉嚐了敗仗，共和黨在眾議院增加了60個席次成為多數黨。政治與經濟變動兩者環環相扣，本研究著重於分析美國失業率對國會期中選舉所產生的影響，探討藉由失業率之數據是否能有效預測並解釋美國國會期中選舉結果。我們僅以時間限制2000年（George W. Bush）至2010年（Barack Obama）的國會選舉作為分析，在本論文裡，我們另外嘗試使用一種獨特的實證方法，那就是運用兩階段的模型—固定效果追蹤迴歸分析(Fixed Effect Panel Regression )和二元評定模型(Binary Logit Model)，而且我們發現研究出來的結果相當好，與我們假設的相關證據一致。除此之外，分析結果顯示，連任率的波動與失業率的變動也呈現一致，保守估計當所有國會選區內的失業率上升2%與4%時，將會降低執政黨3.6%與8.4%的連任率，非常具有參考價值。經過實證研究模型的分析後，顯示失業率的上升，確實會對國會連任機率造成不利的影響。以不加入延遲應變數的實證模型來看，當所有國會選區內的失業率上升2%時，將降低連任率0.036個百分點，當失業率再攀升至4%時，則再降低連任率0.048個百分點；加入延遲應變數後，效果影響甚大，若以所有國會選區內的失業率直接上升4%來看，則國會連任機率將降低0.185個百分點。看似微小的數據，對於政治上的變動可不容小覷，以國會435個席次來看，減少18.5%即減少將近80個席次，已足以影響眾議院內的多數黨。 The 2010 United States House of Representatives elections, Republicans gained control of the chamber, picking up a net total of 63 seats and erasing the gains Democrats made in 2006 and 2008. We also saw the rise of the Tea Party movement ushering in many conservative newcomers to the House. And this was due to the high unemployment rate and bad economic conditions, but why the unemployment rate so important and can make a major change in election? The empirical model we propose to use here is actually a unique combination of two different types of models – a fixed effect panel regression and a binary logit model. In the first stage a fixed effects panel regression model is estimated using the binary variable (1 = incumbent party returned to power and 0 = incumbent party is not returned to power).Hence, the single fixed effects variable described above is then introduced into a binary logit model involving the same set of explanatory variables as the linear probability model in the first stage. By doing so, a set of very specialized effects for the panel in question can be aligned with the logit model and a better analysis of the probability of reelection can be obtained. A conservative estimate shows that a rise in the unemployment rate of 2% and 4% across ALL congressional districts lowers the probability of the incumbent party being returned to office by about 3.6% and 8.4%, respectively. This effect is greatly magnified when a lagged dependent variable is added to the logit model. In such as case, a 2% increase in unemployment across all congressional districts lowers the average probability of return the incumbent party by 7.2%, while a 4% increase in unemployment reduces the probability by an impressive 18.5%.