China Dissertations in France. A preliminary survey (1)
This is the first installment of a series of posts on the production of doctoral dissertations on China in France. My initial impulse was to examine more specifically doctoral dissertations in history, as part of a reflection about the place of modern Chinese history in France (see my previous post on Modern Chinese History in France: A critical appraisal). In the course of collecting data, however, I realized my analysis would benefit from a broader contextualization. The series I begin today will eventually be published as a paper on the PEERS platform with better visualizations and the relevant data. I devote this first post to defining the methodological process and presenting an overview of the survey.
Doctoral dissertations represent a very substantial share of knowledge production in academia. Yet, in the humanities and social sciences many, if not most, never get published and their readership remains limited to the dissertation committee and the scholars who discover these works in the course of their research. Trends in scholarly research are usually examined through scholarly production in journals and books. Dissertations are very rarely the object of reviews as is the common practice for books in academic journals. It is only recently that two initiatives, Dissertations Reviews in the United States and Lisons les thèses in France, have paved the way to support and publicize the review of doctoral dissertations.
In this paper, I wish to offer a different perspective by taking doctoral dissertations as a body of knowledge in itself and explore this corpus through the metadata produced by theses.fr, the official registration platform for dissertations in France. This is a limited approach, of course, because it does not examine the full text of the dissertations, which could be done using advanced techniques of textual analysis and topic modeling. France has maintained for decades a centralized system of registration and microfilming of all doctoral dissertations. The digital turn, however, had the unfortunate effect of preserving only the system or registration. To my knowledge, no attempt was ever made to retro-digitize the microfilms as University Microfilms International (renamed Proquest) did in the United States. Moreover, all the dissertations produced in digital format are now scattered in the universities where they were filed. There is no universal and open access to doctoral dissertations in France.
The purpose of this study is to examine French doctoral dissertations on “China”. I built my corpus from the theses.fr web site. The national platform records all on-going and completed dissertations in all the disciplines in French universities. Because registration is compulsory when matriculating in any Ph.D. program in France, the web site provides an exhaustive record of all dissertations since 1985.
The first question that needs to be addressed is “What is a dissertation on China”? What does “China” actually mean in this context? There is no straightforward way to extract dissertations on “China” because the word “China” can appear in many places in the various fields of the database and because many dissertations simply do not include the word “China” in any of the main reference fields (title, keyword, etc.) of the database. One needs to throw several nets to gather them all, but each net also catches unwanted and irrelevant results.
To build the corpus, I proceeded in three steps: I queried the database for all the dissertations that contained the term “China” in the title. This produced 999 titles. The second step repeated the same query in the keyword field. It garnered 1,000 titles. The last step consisted in using “Chinese” and “mandarin” in combination to identify dissertations in linguistics. This resulted in a new crop of 110 titles. Altogether, I had produced three files with a total of 2,075 titles that, inevitably, overlapped. I did not process further these results} After eliminating the duplicated titles, the corpus came down to 1,610 dissertations. Yet, this included works in unwanted fields of knowledge.
To weed out the irrelevant titles, I proceeded in two sequences of coding. In the first sequence, I reviewed the titles and coded all the dissertations that fell outside of the humanities and social sciences. The use of “China” in the keyword search had produced results about tectonic plates, biology, etc., while “mandarin” injected titles about the zebra finch (“diamant mandarin” in French). In the second round of coding I read more systematically the title to weed out dissertations that may have touched on China, but did not actually study “things Chinese” properly. This was an exercise that definitely involved a degree of subjectivity, but it was necessary to distinguish between dissertations that produced knowledge on China, with or without a knowledge of Chinese, and dissertations where China was just an angle through which to study various topics, mostly in literature (e.g. China in Voltaire’s work).
The final tally at the end of the weeding out process amounted to 1,265 titles. I was almost ready to jump into analyzing the data, but upon preliminary examination, I realized it would require a lot of cleaning before being ready for analysis. The data in various fields reproduce the information as it was entered by the staff at university level. There was no further correction, homogenization, or standardization at higher levels (ministry) to produce usable data. Four fields in particular were problematic: discipline, university (établissement de soutenance), advisor(s) (directeur de thèse), and language (langue). Moreover, the name of authors (Ph.D. candidate), advisors, and the title of the dissertations were written sometimes all in capitals, sometimes with title capitals only, with or without accents, etc. What were the issues?
– Disciplines: the information entered in this field reflects the name of the Ph.D. program under which the dissertation was completed, not the discipline itself. Unfortunately, the metadata does not include the name of the university department which could have helped circumscribe better the production in each discipline. Altogether, I extracted 269 different “disciplines” from the corpus, mostly labels of Ph.D. programs. To homogenize the disciplines, I produced a two-level hierarchy, one with 18 disciplines that still reflected the original programs (though in a much simplified way) and one with 13 disciplines that reflected fields of knowledge. This operation implied a certain degree of aggregation and simplification (e.g. sciences économiques, sciences de gestion, économie, gestion, économie du développement, etc. all came under “Economy”). I maintained “Chinese studies” (études chinoises) in my typology. Even if it is not a discipline, it does express a field of scholarly production that should be identified as such.
– University: the problem here was two-fold. On the one hand, the name of universities was not homogenized (the same university appeared under two denominations) and for Paris several institutions were labelled “Paris, XXXX”. On the other hand, there has been a process of merging among French Universities (e.g. Marseille 1, Marseille 2, Marseille 3 have merged as Aix-Marseille University), which makes it impossible to maintain a statistical continuity among the original universities and the new institution. For statistical purposes, I grouped all the dissertations completed in the individual original universities under the name of the new institution. Furthermore, I extracted and created a new variable with the location of the universities (City).
– Advisors: the issue here was that a number of dissertations was co-supervised by two advisors whose names were listed randomly and alternatively first and second in the same field. There were 130 such cases in the corpus. I kept the original field, but I split and distributed the name of the two advisors in two distinct columns (Advisor 1, Advisor 2). I rearranged their order so that they appear always in the same field.
– Language: information about the language in which the dissertation was written was provided in the form of a code (e.g. fr, zh, en, etc.), with 17 different combinations (e.g. fr/zh, en/fr, fr/en, etc.). By and large, the dissertations were written in French. I homogenized the language based on the title of the dissertation.
Over a period of 34 years, Ph.D. candidates have produced 1,282 dissertations or an average of 38 per year. Yet this average fails to reflect the pattern of constant increase from year to year, which may tell us something both about the interest in China and on the nature of the population of Ph.D. candidates.
Throughout the late 1980s and the mid-1990s, the number of dissertations hovered between 10 and 15 per year, with a moderate push to around 30 until 2007. In the following decade, there was a noticeable upsurge to 40-60 dissertations per year. In the last two years (2017-2018), however, the number dropped again to around 40.
By order of importance, dissertations in economy represent the largest number with 223 titles (17.’ percent), followed by history with 159 titles (12.4 percent), law with 153 titles (11.9 percent), and linguistic with 127 titles (9.9 percent). Altogether, these four disciplines produced more than half of all the dissertations (51.7 percent). Almost at par, we find Art history and archaeology with 106 titles (8.3 percent), Chinese studies with 98 titles (7.6 percent), and political science with 97 titles (7.6 percent). Literature (84 titles), and Geography (72 titles) come close, followed by sociology (64 titles). The least represented disciplines are anthropology (47 titles), philosophy (37 titles), and education (15 titles).
Over time, the distribution changed radically. Economy was quite weak — less than 5 dissertations per year — from the 1980s to the mid-2000s. Then there was a remarkable surge for a decade, then a sharp drop back to the previous level. It is probably too early to draw any conclusion on this drop in the last three years. Law was quasi nonexistent in the 1980s-1990s. There was a movement upward after 2000, but it is only in the 2010s that a substantial number of dissertations were produced. History bumped its way through the 1980s-2000s, and kicked up only after 2010 with a steady increase. In general, after 2010, there was a growing number of dissertations in fields that had previously been stable at a low level, as in political science, sociology and geography. In economy and law, the surge in dissertations is clearly linked to the increase of dissertations by Chinese students.
 Although registration was in place well before 1985, the digital migration of the register starts only with 1985.
 Linguistics here includes: sciences du langage, didactique des langues, phonétique, etc.
 The same search for on-going dissertations produced respectively 426 (title), 75 (keyword), and 19 (linguistic) titles.