Adult education has long been described as an emerging discipline, but there has been little empirical study of its emergence. This study examined 'emergence' by monitoring that particular knowledge base which is unique to adult education. Studies concerned with the theory and practice of adult education are a quantifiable indicator of unique knowledge about adult education. Evidence that researchers in adult education increasingly cite the work of other researchers in adult education would support the contention that the body of knowledge in adult education is growing. The articles published in Adult Education Quarterly between 1971 and 1986 were analyzed using citation analysis methodology. The frequency of citation to previous adult education studies (primary literature) as opposed to citation of studies peripheral to an adult education context (secondary literature) was determined. Distinguishing between citation categories was carried out by analyzing each title cited. The phenomenon of concern in the cited article was interpreted from the words used in the title, and coded dichotomously as 'primary literature' or 'secondary literature'. Each coded item was then recorded under named authors; thus, the cited author was credited for total frequency cited along with the coded category of writing (author of primary literature or author of secondary literature). Reliability measures performed for intra-judge consistency (recoding data), and inter-judge agreement (independent coding of data) resulted in differences in coding of less than four percent for the former, and nine percent with the latter. Validity of the procedures used in coding cited authors was tested by comparing results obtained to a 'standard'. 'Independent experts' were asked to identify from a list of the twenty most cited authors from each four volume period, those who were "primarily known for their adult education activities." The study's coding outcome of these authors compared with the expert's 'standard' resulted in greater than 75 percent agreement. With 4700 citations classified, it was found that a rising percentage of citations were to the "authors of primary literature"; from 41 percent of all citations in the first half of the study period (1971-1978), to 46 percent in the last half (1979-1986). A further breakdown showed the percentage of citations to "primary literature journals" also increasing; from 31 percent of all journals cited in '1971-1978' to 39 percent in '1979-1986'. As the scope of literature analyzed was exclusively from one North American journal, results need to be regarded with this limitation in mind. However, the empirical evidence of an increasing 'primary literature' base in adult education research suggests emergence of the field. Implications for future research are discussed in light of this and previous studies. / Education, Faculty of / Educational Studies (EDST), Department of / Graduate

近些年來，一些論文數據庫(特別是Libra) 變得公開化并已經開始提供給用戶他們請求獲得的信息。這使得我們可以將學術社區當做一個社會網絡來進行研究，不僅分析對作者的著作進行一些統計學的分析，還研究一個作者選擇性與其他作者合著的關係，以及一個作者對其他作者的影響。我們的研究即是定義一些基於社會網絡的方法來測試前面所說的影響關係及合著關係等。 / 我們設計的算法中，最主要的是作者影響力排名算法(AIR) 。該算法類似于著名的網頁排名算法(PageRank) ，並且把我們提取的三種關係都考慮在內。而其他的算法，都是基於某種關係或是某些關係的組合。這些算法包括:聯繫( Connection，利用合著關係、)， 追隨者數量(Follower Count ，利用發表關係)，追隨者(Follower ，利用引用關係)和平均引用數量(Balanced Citation Count ，利用合著關係和引用關係)。 / 這對這些算法，我們設計并研究了一些簡單的特例，通過算法之間橫向與縱向的比較來分析這些算法的特性。在不同的情形下，同一算法的表現并不一致，這是我們引入一個新的變量以便於靈活調整的原因。通過設定不同的變量值，我們利用距離衡量工具來度量這些算法結果的變化。 / 更進一步，我們利用不同的數據集合作為輸入來比較不同算法的表現，并利用一種距離測量工具(Spearman Footrule Distance) 來做算法之間的兩兩比較。在算法的比較中，基於排名值，我們能推斷出關於這些算法的一些結論。而基於累積值的比較，一方面驗證了這些結論的正確性，另一方面也展現出作者影響力排名算法(AIR) 的優越性。同時， 一些來源於現實生活中的排名結果，也可以用來串串證作者影響力排名算法(AIR) 的準確度。 / In recent years, some of the publications database become more publically accessible, and are starting to provide additional information users can query (this is specially the case with Libra). This allows us to study the author community as a social network, analyzing not only the statistics about papers published by an author, individually at a time, but also an author’s choice and extent in connecting to other authors (co-authoring), and an author’s influence on other authors. Our approach is todesign various social network type of metrics to measure the traits defined above. / The main algorithm Author Influence Ranking (AIR), which is analogous to PageRank algorithm, is defined by taking all three relationships into consideration. Other algorithms, based on a single relationship or combination of different relationships, include: Connection, ranking algorithm using coauthor-ship; Follower Count, ranking algorithm using the number of authors who cite papers of a particular author; Follower, ranking algorithm using citation-ship; Balanced Citation Count, ranking algorithm using citation counts normalized by coauthors. / To show properties of different algorithms and do comparison among them, we design and study primitive cases. For some algorithms, the teleportation vector leads to the deviation at certain extent. Therefore, we study different teleportation vectors by tuning a parameter β(details discussed in Chapter 4) from 0 (uniformly distributed) to 1 (proportional to authors’ productivity). With different β, we define Distance to measure the changes in results of these algorithms. / Furthermore, comparisons among these algorithms are conducted by using different publication dataset and we choose Spearman Footrule Distance in our experiment to do comparison for pair of algorithms. Rank value and cumulative value are used in the comparisons: based on the comparisons using rank value, we conclude several observations regarding these algorithms. While the comparisons based on cumulative value help us confirm the "efficiency" of AIR. For using AIR metric, we can find out those really influential researchers who may not be ranked high by other metrics. We study the influence of Turing award winners and all the Turing Award winners scored at least "B", from which we can see AIR’s "accuracy". We also apply AIR metric in the real situation. We study researchers who have Grade "A"(the grade will be discussed in Chapter 6) in Influence and find most of them have good positions in reality, which help us justify the validity of AIR.("efficiency", "accuracy" and "validity" will be discussed more in Chapter 6.) / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Song, Qianqian. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2013. / Includes bibliographical references (leaves 78-82). / Abstracts also in Chinese. / Abstract --- p.i / Acknowledgement --- p.iv / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Motivation --- p.1 / Chapter 1.2 --- Background --- p.2 / Chapter 1.3 --- Our Contribution and Organization --- p.5 / Chapter 2 --- Academic Ranking Algorithms --- p.8 / Chapter 2.1 --- Publication Statistics Algorithms --- p.10 / Chapter 2.1.1 --- Citation Count --- p.10 / Chapter 2.1.2 --- Balanced Citation Count --- p.11 / Chapter 2.1.3 --- Follower Count --- p.12 / Chapter 2.2 --- PageRank-like Algorithms --- p.13 / Chapter 2.2.1 --- The PageRank Algorithm --- p.13 / Chapter 2.2.2 --- Author Influence Ranking --- p.15 / Chapter 2.2.3 --- Science Author Rank Algorithm --- p.17 / Chapter 2.2.4 --- Connection --- p.20 / Chapter 2.2.5 --- Follower --- p.21 / Chapter 3 --- Analysis of Metrics Based on Primitive Cases --- p.22 / Chapter 3.1 --- The Original Case --- p.23 / Chapter 3.2 --- Case for Three General Authors --- p.24 / Chapter 3.3 --- Case for Productive Authors --- p.26 / Chapter 3.4 --- Cases for Productive Author and Coauthor-ship --- p.28 / Chapter 3.4.1 --- Type i --- p.28 / Chapter 3.4.2 --- Type ii --- p.30 / Chapter 3.5 --- Case for Coauthor-ship --- p.32 / Chapter 3.6 --- Cases for Citation Count and Balanced Citation Count --- p.34 / Chapter 3.6.1 --- Type i --- p.34 / Chapter 3.6.2 --- Type ii --- p.36 / Chapter 4 --- Key Parameter in PageRank-like Algorithms --- p.39 / Chapter 4.1 --- The Key Parameter β --- p.39 / Chapter 4.2 --- Comparison Based on β --- p.40 / Chapter 4.3 --- Discussion --- p.43 / Chapter 5 --- Algorithms Comparison --- p.46 / Chapter 5.1 --- The Description of Our Comparisons --- p.46 / Chapter 5.2 --- Similarity Between Different Metrics --- p.47 / Chapter 5.3 --- Two Dimensions Comparison --- p.51 / Chapter 5.3.1 --- Comparison in Algorithms Dimension --- p.51 / Chapter 5.3.2 --- Comparison in Time Dimension --- p.54 / Chapter 6 --- Case Study and Validation --- p.56 / Chapter 6.1 --- AIR v.s Other Metrics --- p.57 / Chapter 6.1.1 --- AIR v.s Citation Count --- p.58 / Chapter 6.1.2 --- AIR v.s Follower Count --- p.59 / Chapter 6.1.3 --- AIR v.s Follower --- p.61 / Chapter 6.1.4 --- AIR v.s Connection --- p.62 / Chapter 6.1.5 --- AIR v.s the First Active Year --- p.64 / Chapter 6.2 --- AIR v.s Rank in Reality --- p.65 / Chapter 6.2.1 --- Ranking Award Recipients --- p.65 / Chapter 6.2.2 --- Top AIR Ranking in Society --- p.65 / Chapter 7 --- Conclusion --- p.76 / Bibliography --- p.78

This dissertation consists of three essays in applied microeconomics on the economics of science. The first chapter contribute to the understanding of fairness and recognition in innovation systems. The second and third chapters study the effect of government policies and university relocation on science and education outcomes respectively. The first paper, coauthored with Zihao Li, studies gender difference in innovation recognition using patent citations. We propose a method to quantify under-citation, by constructing a “should-cite” list for each of over 1.5 million patents based on textual similarity, using state-of-the-art natural language processing technique. We find that female-authored patents are approximately 12% more likely to be under-cited than male-authored patents. Additionally, male inventors are far more likely to under-cite patents written by female inventors. Our findings are consistent with the testable implications of taste-based discrimination but not statistical discrimination. Welfare analysis shows that past under-citations negatively impact future patenting activities, especially for female inventors. The second paper, coauthored with Elliott Ash, Mirko Draca and David Cai studies the impact of a large-scale scientist recruitment program – China’s Junior Thousand Talents Plan – on the productivity of recruited scholars and their local peers in Chinese host universities. Using a comprehensive dataset of published scientific articles, we estimate effects on quantity and quality in a matched difference-in-differences framework. We observe neutral direct productivity effects for participants over a 6-year post-period: an initial drop is followed by a fully offsetting recovery. However, the program participants collaborate at higher rates with more junior China-based co-authors at their host institutions. Looking to peers in the hosting department, we observe positive and rising productivity impacts for peer scholars, equivalent to approximately 0.6 of a publication per peer scholar in the long run. Heterogeneity analysis and the absence of correlated resource effects point to the peer effect being rooted in a knowledge spillover mechanism. The third paper studies the long run effect on local education outcomes of the temporary exodus of Chinese universities in the Second Sino-Japanese War (1937-1945). During the war, over 80% of China’s universities, along with the top tiers of China’s educated talents were forced to relocate to inland underdeveloped areas during the war. We find that the large inflow of educated elite intellectuals and universities increased local supply of secondary schools by 6.6% during and after the war period, indicating the effect cascades to lower tiers of education. However, such trend does not persist into the People’s Republic of China (PRC) era and we find limited effect on local education outcomes in the long run. We discuss the salience of locational fundamentals and education policies in explaining the absence of persistence.