Bassino: Now, I would like to make a report based on my discussion paper entitled "Estimating population and labor force in Vietnam under French rule (1900-1954)." This paper was written in collaboration with a fellow researchers at my university. I primarily handled the data and analysis, and the population estimations for the beginning of the 20th century are primarily the work of Maks Banens's research.
[Report]
Population Statistics for Vietnam
Bassino: Population statistics are, of course, very important to the "Long-term Asian Economic Statistics Database Project". Population has important significance to the analysis of productivity, such as per capita agriculture production. There are almost no official report on the labor force population for Vietnam. As such, the labor force cannot be estimated without using population data.
However, a census of the population of the Indochina area certainly does exist, and we worked toward the goal of developing, as much as possible, estimates of the overall population and labor force throughout the 20th century, using readily available data and recent data on the Vietnam population.
Odaka: By census, do you mean the population census taken by the Vietnamese government officials using the Vietnamese language?
Bassino: Yes. The first census covered only Cochichina (Southern Vietnam).
Odaka: In what year was the census taken?
Bassino: It was in 1901. This first census covered only southern Vietnam. Census reports were made in Vietnamese and French. However, we are not using these reports. Other researchers also do not seem to be making much use of these reports.
Odaka: Were there other censuses after that?
Bassino: Yes. but we have also other sources such as the registration of Catholics (a registration of baptisms and burials). Catholics accounted for nearly 10% of the population, or perhaps less than that, and were spread throughout Northern, Central, and Southern Vietnam. Therefore, by using this data, it should be possible to establish vital statistics (data related to birth rates and death rates) starting with the beginning of the 19th century, but we are not using this type of data, either.
Odaka: That would also prove to be interesting research material. Are the records still available?
Bassino: There are still records in Vietnam and at the Vatican. The records at the Vatican are the more accessible.
Odaka: Using church records, we should at least be able to understand the birth and death rates, right?
Bassino: That should be true for at least the 19th century.
Odaka: Do you have plans for this type of research inAre you planning something for the future?
Bassino: Yes. I discussed this last year with Professor Osamu Saito, and it was decided that it would be best better to stop at this stagefor the moment because too much time would be involved. We decided that it would be better to proceed with the first stage of that research after at least assembling all of the data for the entire 20th century.
We also do not know if the censuses were complete censuses or not. The quality of the censuses varied with the region. As a rule, the censuses taken in the southern part of the country (Cochichina) are said to be fairly accurate.
Censuses were performed in the southern part of the country in 1901, 1906, 1921, 1926, 1931, 1936, 1943, 1946, and 1951. At the time of the 1951 census, there were large areas of the country no longer under control of the French administration, so the census may not have been taken very carefully. It is particularly unclear whether or not censuses were taken during the war in the northern part of Vietnam since most of the countryside was under control of the Viet Minh underground administrative system.
Censuses and Vital Statistics
I think it would be fair to say that, for Northern and Central Vietnam, estimates of vital statistics are more reliable than census data. However, for Southern Vietnam, we have both census data and vital statistics. The census data is somewhat reliable, so both the government officials of Southern Vietnam and the Indochina government-general evaluated the vital statistics as also being reliable. I think that the best estimates are for Southern Vietnam.
There is, however, the problem of report omissions, and this problem is not limited to just Vietnam. In particular, there is the problem of the under-registration of young women and ethnic minorities. Also, in many cases, people such as maids were not registered by the censuses.
Odaka: Why were young women under counted?
Bassino: This is a phenomenon that can be seen all over the world.
Takahama: When the social status of women is low, they are not seen as having much importance, and may simply be forgotten when a census is taken, or their births may go purposely unregistered because the family felt it was too much trouble to go all the way to a government office and register the birth.
Bassino: A more important point is what was often said at the government-general leveloalthough the censuses for Southern Vietnam (Cochichina) are thought to be fairly accurate, they are not very accurate compared with censuses taken in Cambodia. This is because in Vietnam, various types of surveys were conducted with the help of government officials from the old government system, so there was a traditional feeling that it was not appropriate to investigate family matters. Therefore, at the village level, the officials of the village filled in the census forms. It was thought that inquiring about family information was intrusive.
Odaka: Isn't it true that in Japan, village officials knew who lived where, how many women there were, etc.?
Saito: Yes, it was village officials who compiled population registers for their villages in the Tokugawa era.
Takahama: Still, even if a woman was young, she was considered rather important if she was married and pregnant or old enough to bear children, so such woman would likely have been readily recorded in censuses. However, children too young to be considered part of the labor force, especially girls, were not recorded in censuses.
Odaka: You're talking about infants, aren't you? That may have been true all over the world.
Bassino: Another problem, which may be common to other countries as well, is that there are many cases in which people who were actually living in Hanoi or other cities, but still had relatives in their original farming village, were considered to have been temporarily relocated and so they were registered in the village censuses of their home villages. There may also have been These were cases of people being simultaneously registered in two different censuses. There may have even been cases of children who were registered in a village census at the time of birth but then left the village soon after that with their parents who went to a city to find work.
Data about the overseas Chinese merchants nationals living abroad is said to be reliable for the most part. This is because the French police had a policy of dividing the overseas Chinese merchants immigrants into several groups, selecting one person with high social status from each group with high social status, and then controlling the merchants Chinese immigrants indirectly through the people they had selected. As such, there is said to have been careful control over Chinese immigration and emigration.
When writing this discussion paper, we used data that was somewhat readily obtainable, and also included as much province-specific data as possible. Although such data is not immediately applicable, we would like to use that data in the future to study populations by province and to perform consistency checks of our data and estimates.
By the way, in our next stage of work, I we plan to use the Inverse Projection Method to estimate populations starting with the beginning of the 20th century. I have formulated several assumptions to use when making these estimates.
First, although the censuses were not all that accurate, I believewe assume that the vital rates obtained from the vital statistics from southern Vietnam are reliable. Another assumption we have made is that there were few immigrants from China. Thi, whichs will require a bit more examination, and t. There was some movement of people from the North to the South, but the numbers were relatively small compared with the total population. There was also immigration from the mountainous areas of southern China, but that, too, was at a negligible level. Also, we assume that we can apply vital statistics from southern Vietnam to the entire country to calculate reconstructed estimates for the country.
There is a considerable difference between the census reports and estimated values, but this difference narrows if we use methods such as those just described. Also, I think that censuses done in and after the 1950s can be considered to be reliable. Taking the 1986 census as a standard and then looking back over previous years, it appears that the populations for the 1910s, 1920s and 1930s were larger than those reported in the censuses. This is an important point to consider when calculating the future per capita growththe rate of future GDP growth.
Odaka: So you use the 1986 census because it is the most detailed and most reliable?
Bassino: That's right.
Labor Force Estimations
Bassino: Our final step was to develop a graph of the age-gender population for 1908 to 1993 for use in estimating the labor force.
For the purpose of simplification, here we have taken the labor force population to be people from ages 14 to 60. We then compared the results of this graph to the employment statistics discovered thus far. Employment statistics are unexpectedly difficult to aquire, but we were able to easily find numbers for laborers working in coal mines, public primary education systems, and public health services, and later, numbers for civil servants as well. Of these numbers, we limited our survey as much as possible to non-European populations. There was a complete separation between the Vietnamese labor market and the French labor market, so we felt it would be best better to exclude the French from the population counts.
The French population at that time accounted for a mere 0.2% of the total population. The French worked primarily as government officials and in the military, and there were indeed very few French in the country at that time. There was little increase in the number of government officials from the 1910s to the mid-1940s, but there was a rapid increase in the ratio of people working in educational and health-related fields during since World War I, particularly during the 1920s. There was also a rapid increase in coal mine related workers, particularly during the 1920s. I think we can attribute this to the increased importance of Asian-bound coal. There was a slight further increase in labor force engaged in the coal mining industry in the 1930s, but the growth was modest when seen as a ratio of the total labor force.
This concludes the report. I would like to say in closing that the estimates may require some additional compilation using province-specific data.
Odaka: Thank you very much. I would now like to ask Ms. Takahama to comment on this discussion paper and on today's report by Mr. Bassino.
under French rule:1900-1954
This paper presents preliminary findings on population and labour force in Vietnam under French rule, between 1900 and 1954. Data are extremely fragmentary and their reliability is uncertain.
As a rule, the French colonial administration was not much interested in producing accurate figures on Native population as well as labour force, especially before the end of the nineteenth century. Labour force was not regarded as a reliable indicator for estimating income from productive activities since French administration investigations focused on fiscal issues and labour exploitation by French firms and colonial administration. All those problems make population and labour force estimates a difficult but essential task for research on Vietnam's quantitative economic history.
Here, the sources on Vietnam's population under French rule, in particular, Cochichina (southern Vietnam) where the data quality is relatively reliable, and patterns of population and labour force by official series are assessed. Using an inverse population method that exposes a reconstruction of Vietnam population, then estimate the labour force relying on the assumption of labour participation by age and sex.
The results indicate that a large part of Vietnam population was certainly under-registered. This corresponds to a lower increase rate that usually assumed using official sources by French civil officers. Mortality increase during World War II was very limited in time and areas.
Regarding to labour force, it is possible to compare available data on employment in specific sectors to labour force estimates based on age structure of population and assumption of labour participation by age and sex. It is indicated that a share of male labour force peaked during the late 1920s and early 1930s.
Further research could enhance the accuracy and reliability of these preliminary results, and the first aim could be to produce annual estimates of population and labour force. It might be useful to solve several problematic inconsistencies of data to reduce uncertainty and improve Vietnam's population estimates.
[Comments] Mihoko Takahama
Takahama: First, in regard to population, you spoke considerably about data errors, and I believe that this paper required an extreme amount of work and effortthink that it must have taken a lot of work to deal with these errors. Based on the idea that the vital statistics are correct while the censuses are not so reliable, you selected 1986 as a base and worked backward from there. I think that, in order to compensate for the under reporting of vital statistics, it would be a good idea at the very start to separately investigate the birth and death rates. How did you handle these in your research?
If this type of separate investigation is not conducted, wouldn't using the Inverse Projection Method starting with 1986 result in under estimates? This would occur if there was a discrepancy between the (number of) birth registrations and death registrations.
Odaka: Why would the results always yield under estimations?
Takahama: The results wouldn't necessarily always yield under estimations, but there would naturally be a greater likelihood of under estimations resulting than there would be over estimations.
A birth is a vital event that occurs only for the population sector at age 0, but death can occur at any age. Therefore, when there is a report of the death of a person whose birth was not registered, estimates based on such deaths will result in under estimations. However, omissions in death registrations are also a difficult problem. A death is always registered as a proxy report, so the problem of omissions is particularly noticeable for the deaths of elderly people. There are many cases in which reports of the deaths of elderly people were omitted. Since the death was not registered, the data shows the person who has died as still being alive. As such, the problem of under reporting will vary with age, so this is a factor that must be taken into consideration when using data categorized by age.
In this report, two estimates of death rates are presented for the West and South using model life tables, but I don't really understand the reason for selecting the South. This West, South, North, and East are simply divisions into four models as parts of a global model, using death-related data from various countries.
This South model was produced using death-related data from countries like Spain and Portugal, and a characteristic of this model is an extremely low adult mortality, even lower than in the West model. The model is characteristic in that the adult mortality is lower than in any other models, but the child mortality rate is higher than even the West model, and the death rates are high again for people 65 and older. The West model shows the most generalized death rate pattern, so I don't understand why the South model was selected.
Saito: I think that both the South and West were checked. As a result, the South was selected . . .
Takahama: The report says that the West model had a better fit, which seems only natural. As one would expect, the West model shows the most generalized format, and it would seem that the South has little significance in terms of use in the comparison of the fit of death rate estimations.
Odaka: As far as the West model having a better fit, what kinds of data were examined? Does the model best fit 1986Are you talking about fit of the 1986 model?
Takahama: Yes. It fits is the fit for the estimated values for the reported values of the 1986 population age structure.
Also, in the validation of data errors, we make particular use of the gender-specific birth ratio when looking at birth rates. For humans, 105 boys are said to be born for every 100 girls, so we can easily understand the variation between the sexes. For death rates, we can apply the life table to work out a somewhat accurate death rate, minimizing problems, but I still think that we need toshouldnit set specific corresponding time periods for the gender-specific death ratios.
By the way, it is said the report says that migration data by age, in five-year age blocks, is needed as an assumption. Is such data actually available?
Bassino: We have ignored migration in this research.
Takahama: Ignoring migration also helps make for a better model fit, doesn't it?
You have made two assumptions. One is that, by using both the vital statistics and census statistics of Cochichina to estimate migration and then incorporating the estimates into your assumptions, you arrive at birth rate and death rate estimates that are representative of the entire country of Vietnam. The other is an assumption that there was absolutely no migration. The latter assumption is the better assumption, isn't it?
[Discussion]
Odaka: Let's move on to the discussion.
Bassino: Thank you for your comments.
The first problem is that the censuses are incorrect while the vital statistics may be more reliable. I discussed this with my fellow researchers who worked on this project with me, and we felt that, because French rule in the South was rather strict and well established, vital statistics may have been made for the entire Southern region, while in Central and Northern Vietnam vital statistics were recorded only in the areas that the French had firm control over.
This is indeed a problem. In this sense, it would be helpful for us to have vital population statistics from the Catholic Church. One hypothesis we considered was that the vital statistics were the more accurate and that the censuses were less accurate, and at this stage, I don't think we can assume anything more than that. Conversely, however, according to all reports I have seen, there's nothing to support the idea that it was the censuses that were correct and the vital statistics that were not correct.
Odaka: Of course there is the consideration of what is correct and what is incorrect, but as the censuses were only performed at sporadic intervals, the only thing we can use to fill in the gaps, regardless, are the vital statistics.
Bassino: That is also true.
However, the biggest problem in using vital statistics, even more important than the fact that there are omissions in death rate records, is that these statistics only apply to Southern Vietnam. Of course, around the end of World War II, there was a major difference in the death rates between Northern and Southern Vietnam.
Takahama: By "strict conditions," do you mean that hard penal measures were used?
Bassino: That's right. However, in the case of vital statistics, the population in the South was large and the area was under complete rule as a French colony with French systems introduced, so I believe that there are not many problems with the vital statistics from the South.
Now, in regard to the selection of the West and the South tables, various calculations showed that these two tables had the best fits, and that the West was the better of the two. We used a general mortality table, so I think our results should be acceptable.
Saito: So, with this method of estimation, you considered the vital statistics for Cochichina to be correct, and expanded the statistics to apply to the entire country. You also made an assumption of zero migration. I think that, perhaps, if you're trying to make estimations for the entire country, the use of a zero migration assumption is unavoidable. I too am making the same type of estimations for India, and have to use a zero migration assumption as well.
However, there is also another method that can be used. Since the data for Cochichina is quality data, the Inverse Projection Method can be applied, (but) to the South only. Doing this, we have a problem of migration. The assumption of zero migration is, of course, not realistic, but we can still apply this assumption to our estimations. I think that it would be all right to make estimates only for the South by using vital rates obtained from vital statistics that are comparatively reliable, and by making a strong assumption of zero migration, providing the results should be checked.
As no one could make estimates without a bold assumption in this case, we might as well try expanding the vital statistics for Cochichina to a national level, and or try closing our eyes to immigration and applying zero migration to the South only. I think either method has its faults. Since both methods are faulty, I suggest trying both (and comparing the results).
Odaka: How would the North be handled in that case?
Saito: No estimations for the North would be made for a while. However, if similar results are obtained from the two procedures, the initial results could be left as they are. If the results are very different, it may indicate a problem.
Another suggestion I have is to use other methodologies, not only the Inverse Projection Method. For example, one method we are using with the population of India is the growth balance method. This method makes calculations using the assumption of yearly stable population. There are some extremely unrealistic aspects to this method, but one benefit of this method is that it automatically calculates correction factors and the completeness of registration.
We are studying a specific region of India, so there is no way that we can make the assumption of zero migration. Still, looking at results, the completeness of registration was stable at around 80% throughout the period from around 1910 to the end of the 1930s. Considering migration, this is exceptionally good. However, this ratio dropped slightly prior to that period. But because we know this will occurredthese facts, I believe it is worth trying various methods. There are several estimation methods available, such as the Logit Model, so it would be worth trying different things.
I don't think it would be a good idea, though, to attempt calculating estimates for the entire country without paying attention to alternatives. After all, we know that the data is inaccurate. So, it would probably be best to start by making estimates with various methods for the South only, since relatively accurate data is available.
Actually, we first used a different methodology in our study of India, and we may end up deciding to try the Inverse Projection Method. That would be the reverse order of what I've just suggested.
Bassino: The final goal of this project is not to cover Indochina, but the Vietnam, so Pr. Saito's suggestion is a very good one for this project.
I have also given this a lot of thought. I thought that it might be sufficient just to use Cochichina, but we only treated Cochichina separately in terms of population series. For other data we will deal with Vietnam as a whole, not Cochichina.
Takahama: I understand that there are time limitations, so I think it might not be possible (to try many different methods). Still, if a few other methods are tried and the results agree, you could say with confidence that the results are indeed representative of Vietnam.
Bassino: It may be a good idea to try a two-stage method. First we try with Cochichina, and if the results are good, we then expand the scope to include the entire country.
Saito: I believe there are two possibilities. Let's say that you can get somewhat acceptable results for Cochichina. Next, when estimating the total population for the entire country, if the same estimation method does not yield good results, proportional distribution of population can be applied using a census from a specific period, say 1920 or 1921. The scope can then be expanded to the entire country based on the Cochichina percentage. This is one method that I think is applicable. It is a very rough method of estimation, but I think it is equally possible to reach results that are fairly acceptable for the key Southern area and also yield results for the country as a whole.
Bassino: In regard to the point proposed by Dr. Saito, another thing we have to consider is that migration was not negligible in the South. We do have most of the statistics for migration from Northern and Central Vietnam to Cochinchina, and immigration from China, so we will try adding these to our estimates.
Saito: Still, there are no detailed profiles of the migrants, are there?
Takada: In regard to censuses in the North, a strict system for imposing poll tax in the North was implemented just prior to the 1920s. Before this, it was left to administrators in each village. The (government) had only a (general) idea of the number of people in each village, but in terms of economic matters, a large amount of money had to be collected for the poll taxes, so the government felt for the first time that it needed to know exact populations. Therefore, just prior to 1920, the (government) started taking accurate counts of the adult males in the villages.
So, although the statistics for 1921 contain population counts by gender,
these figures may have been estimations based on the average number of
people in a family and the ratio of men to women in a family. It is difficult to know whether these figures reflect actual conditions or not.
There would have been many people in each village that would not have been part of a particular "family", so I think there would have been actually more people in each village than the statistics show. I think that the statistics for the North for the 1920s and later will end up being about the same as those for Cochichina. Even for Cochichina, the "censuses" were probably left to each village, with the village officials creating new figures for each village based on the poll tax. We can hardly think that the censuses gave accurate results of the number of people in each household who were to pay the poll tax.
The South was in a state of social fluxothis was an area undergoing pioneering development, so there was a high level of migration. Therefore, we can expect that it would be difficult to accurately determine the population (at any given time). Even if we had accurate population counts for the South, it would be somewhat meaningless to estimate the population for Vietnam as a whole based on such figures. More than that, even if we looked at the North using censuses, as is, from the North taken in the 1920s and later, we would still probably end up with nearly the same results.
Bassino: With the North, there is the problem of ethnic minorities in the 1920s and 1930s.
Takada: Censuses from the Yagantze Huanghe (Rouge River) Delta are more accurate. Later, I would like to show Mr. Bassino how the numbers for village poll taxes were recorded.
Takahama: The reason behind a collection of statistics also greatly affects the actual data records and the reliability of the data. For example, many people would have avoided censuses to prevent being drafted by the army or having to pay taxes. So, if a census is taken to determine the number of people in a family for which taxes are to be paid, the numbers reported would probably be lower than the actual numbers. Still, this is a general theory, and I don't know if it would really apply to Vietnam.
Tran: My comment is about the estimation of labor force. You say here that anyone 15 and older is considered part of the labor force, and anyone younger than that is negligible as part of the labor force. This may be applicable to modern-day Vietnam, but I think this method of age division may not be applicable to earlier periods of Vietnam's history, particularly prior to World War II.
I lived in central Vietnam as a child. There were still no very few modern elementary schools at that time. Nearly all girls learned to read and write at home, being taught by older brothers and sisters. Even many boys were unable to go to elementary school. Some boys might go to something similar to the private elementary schools of Edo-period Japan until around the age of 10, and then an extremely small number of those might go on to a school in the city.
Even today, approximately 75% of the population lives in farming villages, but approximately 90% of the country's population prior to the war lived in farming villages. I therefore think that most children, even under the age of 15, would have been performing some kind of labor. So, I would think that Mr. Bassino's assumptions related to labor force would result in under estimation.
Bassino: I also thought it would be better to count children as young as 10 in the labor force, but according to Dr. Saito, unlike poverty-stricken families in today's large cities, in a traditional agrarian society, work done by children from the age of 10 to about 14 was considered to be a type of apprenticeship, so we decided to set the cutoff at 15.
I am not so familiar with traditional rural life in southeastern Asia, and Vietnam in particular, but, for example, if we consider the traditional Mediterranean societies of around the end of the 19th century, children up to the age of 14, even if they did work such as taking care of sheepgoats, were not considered to be full-time laborers, but rather, assistance or apprentices.
Tran: But, the majority of the work was agricultural. The children were not working in industries, they were working in service areas or in agriculture.
Bassino: Of course that is correct. Still, there are various types of apprenticeships, even in agriculture, and we felt it would be better to choose 15 as the cutoff than age 10.
Tran: While I think it is true that the children did not perform exactly the same type of work as adult laborers, I also think that the labor population under age 15 accounted for a large percentage of the total population. This is true even today, so the ratio should have been even higher in earlier years in Vietnam. Also, there is an extremely low percentage of children advancing on to higher grades of school. As such, even if the children did not perform exactly the same type of work as adult laborers, I think the value of the labor performed by youth is not negligible.
Bassino: We are currently collecting materials and data related to education. We developed estimates of the labor force basically to measure the ratio of the labor force to the total population. What we ultimately want to study is the ratio of laborers working in the coal mining industry, in particular, compared with the total labor force of the 1920s and 1930s.
Odaka: We have had various comments and opinions, and would like to ask Mr. Bassino to take these into consideration when revising his paper. I would like to thank you all for your participation.
(3rd August 1998, at Institute of Economic Research, Hitotsubashi University)