A Spatial Analysis of College Choice Behavior and Its Regional Difference in Taiwan

[[abstract]]近十數年來台灣地區大學院校與招生人數大幅增加，招生競爭之激烈已成事實。因此，大學入學選擇決策的相關研究，對於各大學院校招生行銷策略的擬定，將可提供有價值的參考資訊。大學入學與選擇於1970 年代已為許多美國的社會學家與經濟學家所關注，並擴展至高等教育研究者、大學經營管理者以及其他社會學科的學者。從地理學的角度而言，大學屬性包含了地理空間要素，因此大學入學選擇可說是學生面對眾多空間選擇機會所做的決策行為。這也是國內大學入學選擇的相關研究中，不曾出現的研究觀點。本文將以Rushton 的空間行為研究架構，結合已廣泛運用於許多學科領域的隨機效用模式，分析台灣地區大學入學的空間選擇行為及其影響因素，並據此推論大學入學選擇行為的區域差異。根據隨機效用模式，決策者的選擇結果可視為其對於各個替選方案偏好程度高低的反應，並且偏好程度的高低受到各個替選方案屬性以及決策者本身屬性的影響。偏好資料的取得可有顯示性偏好與敘述性偏好兩種方法，由於資料取得的限制，本文利用敘述性偏好法，亦即以實驗設計的方式設計問卷，考量各大學學雜費、公私立別、學生人數、距家遠近、學校所在地區的人口數等大學屬性，以及性別、學習成就、家庭社經狀態等學生之家庭或個人屬性，進行大學入學選擇的偏好調查。問卷發放以分層比例方式，針對台灣本島地區的高三學生進行調查。所獲得的問卷調查結果，則使用多項羅吉特模式（MNL）校估各模式參數。參數之校估與檢定結果顯示，大學屬性中空間要素－距家遠近以及學校所在地區的人口數確實影響高三學生的大學入學選擇；同時，學生的偏好結構呈現出明顯的區域差異。[[abstract]]During the past ten years of so, colleges and universities in Taiwan have experienced an increasingly intensive competition among one another in recruiting students. Researches in college choice behavior of high-school students could have provided valuable information to the recruitment policies for college administrations. Issues on College attendance and college choice have received scholarly attention from sociologists and economists in 1970s. Other social scientists, researchers of higher education and college administrators have also developed research interests in the college choice processes. From geographic point of view, college attributes include certain geographical elements. College choice behavior, therefore, can be regarded as the result of decisions made by students among various spatial alternatives. This point has rarely been explored in similar researches in Taiwan. The purpose of this research, based on Rushton’s conceptual framework of spatial behavior and combined with random utility models applied in many disciplines, is to analyze college choice behavior of high-school students and its regional difference in Taiwan. According to random utility models, the decision made by an individual is regarded as the response to preference ranking of alternatives which is affected by related attributes of alternatives and of that individual. Due to the unavailability of revealed–preference data, this paper interviewed students through stated preference questionnaire to collect relevant information on (1) college attributes including tuition and fees, public or private institution, total number of students enrolled, distance from home location, population size of college location, (2) student (or household) attributes including gender, academic ability, family socioeconomic status, and (3) reported choices of students. A stratified random sampling scheme was used in this study. These data were then used to fit multinomial logit models (MNLs) to obtain parameter estimates. Results of this analysis showed that spatial attributes of colleges do affect college choice and that students’ preference structures have significant regional differences