I. Sources on Vietnam's demography under French rule

Regarding time series on Vietnam's demography under French rule, Indochina's statistical yearbooks are the main sources used at the present stage of research. Further investigation in the Archives of Aix-en-Provence will be needed in order to complete several series.

For the study of the Vietnam population under French rule (from 1900 to 1954), many problems appear as for the available data's nature and their reliability. The conditions of data production vary very much for each case, making a common interpretation difficult. This heterogeneity of the colonial sources, which also concerns the other French colonies, poses a problem of size: are we always measuring the same phenomenon in the same way? The administrators' choices in time and space make the researcher's task difficult. However, we may venture that the variability of sources as well as their value (their reliability or distant approximation) express the nature of the relationship between the administrators and the administered population.

1) Population censuses and estimations

Until the real first census considered as almost reliable, and which took place in 1979, it is difficult to have a clear idea of the Vietnam demographic situation. The absence of reliable data, the "chronic" under registration, the discontinuity characterising several series, the turbulent history (70 years of colonial rule and 30 years of war) and many other problems have led to an heterogeneous and fragmentary information, which is difficult to appreciate and to handle. In spite of the improvements recorded in the collection of data from the beginning of the century until the fifties, the obtained series only constitute very gross approximations. Thus, if many observers have been able to provide various figures, they had no precise idea of demographic trends before 1945.

In 1901, under governor Doumer's administration, the first "test" census took place in Cochinchina, a French colony. Subsequently, the results of an estimation implemented in 1906 for all Indochina were published as "statistic of Indochina population" in the Official Gazette (August 1907). Vietnam population was then estimated at approximately 13 millions. A new census took place in 1921 in continuation of the 1901 one (these are the only real censuses, using reports for each household). The results seemed satisfactory, especially in Cochinchina, because of a better receptiveness to the administration's requests. In the main cities, however, a significant part of the population appeared to be missing. The 1921 census took place in Cochinchina, but also in the protectorates of Tonkin and Annam. However, in these territories, local administration was free to choose its methods. This introduced heterogeneous techniques within Vietnam.

In Tonkin, no generalised use of reports was made to count the population of each house or each boat ; most of the time, it was the chief of the village who proceeded more or less seriously to the evaluation of the population of the village. Besides the approximate character of counting, many problems appeared, such as the preferential registration of Natives in their native village rather than in the place where they actually lived. In Tonkin and Annam, local administration decided to use arbitrary coefficients in order to correct the population figures, which caused one more major difficulty. Thus, for Annam, the crude figure of 4 183 000 residents in 1921, too low compared to the previous estimation (- 1 500 000), has been boosted to 4 933 000, assuming that salt consumption per capita was the same over the three Vietnamese regions (salt consumption data were recorded by public monopoly, and Cochinchina population estimate was considered as reliable).

These coefficients were abandoned in 1931 and a survey "as precise as possible" was done in all of French possessions. But it was not satisfactory despite the effort to homogenise the results. The same process occurred in 1936, 1943, and in 1946 (for the Cochinchina). In 1951, other population estimates were made for Vietnam, based on approximations provided by local authorities. Moreover, (as, for example, in 1951), only the "controlled" provinces were concerned by these estimations. We may have a doubt on the reliability of the figures, especially for Annam and Tonkin. Even if several series seem coherent, caution remains necessary in so far as evaluations were based on the previous ones, regardless of changes in methods and data's quality.

Because of the use of coefficients correcting crude data and of wrong data as benchmarks, Annam presents the most doubtful demographic data. The problem however appears quite critical for Tonkin too, where whole categories of population were missing (forgotten) at the census (Europeans' servants and their family in 1931 for example). It is obvious that the great number of problems met (approximations, under declarations, insecurity of the country,c) call for criticism. According to Smolski, a demograph and statistician working in the Indochina civil service, margins of error exceeded 50% in several provinces of Annam, between 20 and 25% for Tonkin and about 10% for Cochinchina.

2) Civil status and vital rates

The study of vital rates is based on statistics recorded by the civil status but, again, data reliability is not the same in Cochinchina as in Annam or Tonkin. In Cochinchina, native population civil status appeared as early as 1883, in consequence of a decree that introduced compulsory registration of births, weddings and deaths by local authorities. Despite of hard penal measures (in case of absence or misrepresentation) and of a control by authorities, the figures provided were not entirely satisfactory and do not account for regular kept registers in all of the provinces.

In Tonkin, an outline of a civil status appeared only in 1906. But again, data cannot be considered as reliable, partly because the local administration was slowed down the application of the civil status for earlier mentioned cultural reasons (Confucian value considering this registration as an intrusion within the family). Registration became compulsory in 1924, allowing real improvement. Much later, in 1931, registry offices appeared in Annam, giving however feeble results during the first years.

More generally, several problems appeared at the time of the population's registration :

- inscriptions of individuals in native villages, inducing an important underregistration in the cities.

- girls' underregistration, but also, in some degree, boys' underregistration, especially among ethnic minorities.

- unpublished series, because of very important mistakes and gaps, mostly regarding uncontrolled areas, where births were impossible to check.

- loss of many documents and records, in consequence of the Japanese invasion, and thus of the formal end of the French administration, in March 1945.

However, in spite of these limitations, most of the series on Saigon and Cholon, and more broadly on Cochinchina, offer reliable vital statistics. In addition, Catholic missions, by their involvement in ceremonies such as christenings, weddings or burials, provide complementary information on vital rates that can be used to improve and reconstruct data on Annam and Tonkin.

3) Problems related to administrative division and urban population

In addition to the quality of the statistical sources, which remains to be controlled, it was necessary to solve other problems, especially those concerning the administrative division of Vietnam. Indeed, it is interesting to observe statistical time series, not only for all of the country but also for provinces, where some population information exists.

The major obstacle comes from the numerous different divisions in time. We observe many name of border alterations which make longitudinal analysis difficult. Therefore, it is first of all necessary to institute tedious regroupings so as to give a certain continuity in time and to enable comparisons of provincial series. Besides, from 1954 to 1975, Vietnam was cut into two parts by the 17th parallel, established as "a temporary line of military demarcation", passing through the province of Binh Tri Thien. For all these reasons, we can not simply use population censuses of the late 1950s to estimate the population in uncontrolled areas between 1945 and 1954. But we hope to find, in the Vietnamese archives, population data, recorded in Viet Minh reports of this period.

For the urban population, our research relied upon data mentioned in the statistical Yearbooks but also on those collected by Brenier (1914) and by Mitchell (1995). In this paper, we present only urban population in Hanoi, Haiphong, Saigon and Cholon. It seems that populations of neighbouring suburbs or villages were incorporated following successive changes occurring in the administrative area of these cities. Additional data were found on the population of Hue, Da Nang, Dalat and other smaller towns. Political unrest during WWII and the Viet Minh uprising after 1945 induced massive move of European and native population in towns controlled but the French administration, but these refugees were not necessarily registered.

4) Labour force and employment

Estimation of labour force in Vietnam under French rule is a very difficult task. The colonial administration took an interest in these figures only for undirect motives : control of the Chinese population, of the effective migration flows from Tonkin and North Annam into Cochinchina, of the effective production in mining, using labour force as a proxy for factor utilisation and assuming that labour productivity was roughly constant. For these reasons, there is no comprehensive census of the labour force except for the employment in a few sectors, such as mining, railways, telecommunication and more broadly the civil and military administration.

In order to estimate the labour force, a first approach could be to measure the population older than 15 and younger than 60 or 65 years. Given Vietnam's cultural and social context, a mainly agrarian society with a strong emphasis on education, we can assume that the labour force under 15 years was negligible. Of course, young boys and girls contributed to different domestic tasks, and to agricultural and other productions, but they had a status of part time apprentice rather than full time worker. This estimate, however, will be based on the age structure obtained by the population reconstruction.

A complementary approach would be to measure the labour force in the different sectors and aggregate all these series. This is rather easy for the employment in public sector (civil administration, Native Guard), and for education and health services (private and public, but excluding military medical services). These series are presented below. For the private sector, data on employment in mining are remarkably accurate for coal mines, and to a lesser extent for different metal ore mines. Further research on private archives should bring some light on employment in French or foreign - owned companies. We mentioned as a benchmark, the estimate for 1939.

But this represents only a very small part of the production in manufacturing, cottage industry, construction or services. Surprisingly, given the amount of works on colonial exploitation by historians, labour force fluctuation remains unclear. In addition, the share of Chinese or Vietnamese enterprises has not been considered, until now, as an important issue for the research on Vietnam's quantitative economic history. Estimation procedure of these figures could use, on the one hand, private Chinese and Vietnamese archives on consumption and production, when these series are available.

Estimating labour force in agriculture, forestry, and, if possible, in fishing and hunting, is the most difficult task. Except among plantation workers (rubber tea, coffee) and Chinese fishermen, it was mainly a self-employed population. Estimation will necessarily use population data by province, which are, again, quite unreliable. Regarding agriculture, we may consider harvest data and on reports by the colonial administration on labour productivity in the different areas. But this kind of procedure requires to take account of the natural conditions. Regarding fishing and forestry, we could use fiscal sources to estimate labour force through controlled production (salt monopoly and forest administration) but a significant part of Natives' production (excluding therefore Chinese) was fresh unsalted fish and charcoal, sold locally, or even, for both items, part of the self-consumption. Therefore, the estimation of the labour force, engaged in these productions, as well as in hunting, should be based on works by specialised anthropologists.