本文使用「台灣教育長期追蹤資料庫」第一、二波之學生、家長與教師的問卷資料,對 學業能力的「同儕效果」分別利用最小平方法、固定效果模型以及工具變數模型來進行估 計,並考慮「常態編班」與「能力分班」對於係數估計的影響,說明在不同編班制度下,同 儕效果的程度有何異同。經過內生性的排除,我們發現在總分為 100 分下,同儕每多 1 分, 學生個人也會額外增加 0.192 分,且能力分班下的影響程度較常態分班大。此外,本文也進 一步探討「性別」、「居住地」與「公、私立學校」的同儕效果差異,發現「男女」的同儕效 果係數分別為 0.183 與 0.196 分,顯示整體差異不大,也挑戰「男女分校」的傳統思維;另 外「城鄉」學校的同儕效果係數為 0.227 分,明顯優於「都市」的 0.165 分,推測生活圈和 家庭的連結促使學業的整合性較高;而「私立」學校由於學風嚴格、學生間具有革命情感, 同儕效果為 0.343 分,遠高過「公立」學校的 0.182 分,證實同儕效果存在非線性影響。
This paper used the questionnaire data of students, parents and teachers of the first and second waves of the ”Taiwan Education Panel Survey” to estimate the ”peer effect” of academic ability by the least squares method, the fixed effect model and the instrumental variable model respectively. Considering the influence of ”normal class assignment” and ”ability class assignment” on the coefficient estimation, it shows the differences in the degree of peer effect under different class assignment systems. After the exclusion of endogeneity, we found that under the total score of 100, for every 1 more points earned by the peers, the individual student will also gain an additional 0.192 points, and the impact of the ability class is greater than that of the normal class. In addition, this paper also further explores the peer effect differences of ”gender”, ”place of residence” and ”public and private schools”, and found that the peer effect coefficients of ”male and female” are 0.183 and 0.196 points respectively, showing that the overall difference is not significant, and challenging the traditional thinking of ”separate schools for boys and girls help study”. In addition, the peer effect coefficient of ”rural” schools is 0.227, which is significantly better than the 0.165 points of ”urban”. It is speculated that the connection between life circle and family promotes higher academic integration; and Due to the strict study style and revolutionary sentiment among students, the ”private” school has a peer effect of 0.343, which is much higher than the 0.182 point of the ”public” school, confirming the non-linear effect of the peer effect.