Danish Polar Center | Strandgade 100 H | DK-1401 Copenhagen K | Denmark
phone +45 3288 0100 | fax +45 3288 0101 | 
News    |    Research & Logistics    |    Publications    |    Library    |    Photos    |    Polarfronten    |    About DPC
You are at:

Ethnic and Demographic Aspects of the Arctic Living Conditions Project

by Thomas Andersen and Jette Jensen

1. Introduction
During 1997 Statistics Greenland initiated a circumpolar study of living conditions among Inuit and Saami peoples. The overall aim of this paper is to present this project and two crucial aspects of the study to a broader forum. In section 2 of this paper we briefly introduce the project.
As the project focuses on specific ethnic groups, valid measuring of individual ethnic ties is a crucial element of this project. In section 3 we discuss the theoretical definitions and empirical operationalisations of ethnic groups to be able to distinguish Inuit and Saami from other inhabitants of the areas covered by the survey.
In section 4 we deal with two demographic aspects of this study: Fertility and mortality (including infant and adult mortality). These are important measures in several ways. Firstly, both fertility and mortality reflect the economic, cultural and social status of a given population group, i.e. economic, cultural and social factors have a strong influence on fertility and mortality rates. Therefore, these demographic variables can be seen as indirect measures of living conditions. Besides that, both variables are indicators of health which is an important aspect of human living conditions. The mortality rates are direct measures of health while fertility is an indirect health indicator.

2. The Arctic Living Conditions Project
The Inuit and Saami populations in Inuvialuit, Nunavut, Nunavik, Labrador, Greenland, Alaska, Chukotka, Norway, Sweden, Finland and the Kola Peninsula have a number of economic, cultural and technological conditions in common. This is especially true for the traditional occupations on which these populations still base their livelihood making them dependent on natural resources.
However, the basis for this traditional lifestyle is changing, and these indigenous populations may thus be said to find themselves in a transitional phase between traditional and modern society. The fact that these population groups have had to change their lifestyle within a relatively short span of time has not been without its costs. Indicators show higher unemployment, lower income levels, poorer health and more social problems among Inuit and Saami than among the rest of the population in these countries. As a result, it is important, from the perspective of policy planning as well as research, to be able to document the present level and any future changes of the conditions of life in these areas. Research on the living conditions among indigenous people in the Arctic has, however, been sporadic so far, and, at best, has been dominated by regional studies. Thus, a distinct need for a comparative investigation of the individual living conditions in this area exists. In this connection it is very important that a new research design is created including both the development of a battery of indicators for the living conditions, which have been adapted to the specific economic and cultural characteristics of the indigenous peoples, and a statistical method for sampling in the Arctic.
The expected outcome of the project will be:
1. The development of a new research design for comparative investigations of the living conditions of the Inuit and Saami populations in the Arctic. This will include the drawing-up of a battery of nominal and operationalized indicators of living conditions as well as the development of statistical tools for transnational analyses of living conditions in the Arctic.
2. A mapping of the living conditions among the Inuit and Saami in Greenland, Inuvialuit, Nunavut, Nunavik, Labrador, Alaska, Chukotka, Sweden, Norway, Finland and the Kola Peninsula, which will facilitate inter- and intranational comparisons of the level of the living conditions in a number of dimensions.
3. An improved basis for decision-making in relation to policy planning and implementation.
4. The establishment of a net of contacts between researchers and research institutions interested in societal research in the Arctic.
2.1 The theoretical need
In 1994 Statistics Greenland carried out the first study of living conditions in Greenland since the introduction of Home Rule. As research within the field of living conditions was generally at a pioneering stage in Greenland, the research design was very much a copy of the Scandinavian model for such studies as were the living conditions indicators which were those generally used in the Scandinavian tradition.
The model may be described in the following points:
The Scandinavian model is closely bound up with the policies of welfare and planning. This means primarily that the focal point is the question of differences in the distribution of benefits among the various population groups in society.
When selecting the relevant welfare components, the resources of the individual are emphasized but not the satisfaction of his/her needs or preferences.
When measuring living conditions, the primary indicators are descriptive rather than evaluating or subjective ones.
The view of society and the methodological assumption in the model imply that it is meaningful to speak of a society as a whole which, across class differences and conflicts, strives for a common goal.
It is presupposed that it is possible to talk about universal or generally necessary within the period of time to which the study refers.
The focus is on poor rather than good conditions.
The definition and measuring of welfare/living conditions are based on the idea of the individual as an entity with material as well as nonmaterial resources.
Even though the Greenlandic survey was inspired by the Scandinavian approach, the questionnaire primarily contained material questions. This was due to the lack of basic statistics in a number of areas.
In connection with the analyses and reporting of the Greenlandic living conditions study 1994, it became obvious that problems arose, when a research design presupposing a high degree of social and economic homogeneity internally, and hence the presence of universally necessary resources from a relatively homogeneous late industrial society, were transferred to a country which is first of all characterized by a nonparallel development, secondly can hardly be described as industrialized and thirdly has its own cultural and historic profile. The focus on material indicators only increased these difficulties.
In practice, two major problems followed from using the Scandinavian research design and focusing on material welfare dimensions. Firstly, as one could expect, the inhabitants of the settlements (approximately 1/4 of the Greenlandic population) scored very lowly on traditional indicators such as level of income, housing facilities, level of education, employment etc. Secondly, some of the questions, especially regarding income, property and employment/unemployment, did not result in valid answers.
2.2 The theoretical and methodological implications
The fact that the people living in the Greenlandic settlements in general scored lowly on most of the living conditions measures indicates that the underlying concept of welfare and thereby the welfare dimensions employed in the survey were inappropriate. The obvious question is of course why people are staying in the settlements if they are suffering from bad living conditions. There are some structural limitations such as unemployment among unskilled workers and housing shortage in the cities which to a certain degree prevent people from moving there, but the fact remains that people and politicians in the settlements actually fight for their right to live in the outer areas of Greenland and to maintain the settlements they live in.
This problem calls for a revised and expanded concept of welfare adapted to the reality of the Greenlandic population. When the homogeneity assumption from the Scandinavian model is not present, the battery of living conditions components should be reconsidered with a view to detect the intersocially different welfare priorities. Or expressed in other words, to investigate which welfare components are salient for the different population groups.
The problem concerning invalid questions indicates, however, that a new research design should not only focus on finding proper welfare dimensions but also on reoperationalizing traditional living conditions indicators such as level of income, property and housing facilities in a meaningful way.
These experiences indicate a need for the development of a new research design for the study of living conditions in countries/regions which from an economic and, especially perhaps, a cultural point of view is decidedly different. This is the case for all the countries where Inuit and Saami peoples are living in the Arctic. Only in a few cases, however, have research designs been used in the general research on living conditions, which take into consideration the specific lifestyle and the general dissimilarity of the Inuit and the Saami peoples to the majority of the population.
It is the theoretical and substantial aim of this project to develop such a research design for a comparative study of living conditions among Inuit and Saami peoples in the Arctic.
2.3 The concepts of welfare and living conditions
The overall aim of surveys of living conditions is to measure welfare at the individual or household level. Welfare is a multidimensional concept that varies with time and space. It is one of the main objectives to define and operationalize the concept of welfare in an adequate way. It means to identify the most crucial welfare dimensions in the Inuit and Saami societies today. It is an underlying assumption of the project that this can be done by means of empirical research.
As regarding the concept of welfare, living conditions have been defined in several ways and there exist different related concepts. Below we have described a holistic model that shows how living conditions will be perceived in this project. Focus will be on material resources as well as on nonmaterial resources, and objective as well as subjective living conditions indicators will be applied in the survey.
The concept of living conditions
Figure 1
Usually the concept of living conditions is further defined by the fact that the survey results regarding individual well-being are combined with basic socio-economic statistics. This is also the case in our definition of living conditions. The reason for this is that the satisfaction of needs and the employment of individual resources to a certain degree is structurally determined. In this connection it is relevant to develop and apply the arena approach which has its origin in Coleman's social theory and which was included in the first Norwegian study of living conditions at the beginning of the 1970s. The point of this perspective is that living conditions cannot be defined only as the individual possession of a number of resources - but as the individual possession of different resources which can be brought into action in different arenas (the labour market, the private sphere, political/public life, etc.).
The model is of course rather structural and deterministic, but if modified it can illuminate the fact that the economic, political and geographic structures and the collective outputs (economic growth, political efficiency etc.) from society influences the needs and the employment of individual resources.
3. Ethnicity
Contrary to common studies of living conditions this project focuses on specific ethnic groups - and in most of the areas national, ethnic minority groups. One of the greatest challenges is therefore theoretically and empirically to distinguish between Saami/Inuit peoples from other parts of the populations in the areas in question. This is a problem because there is no appropriate statistical information regarding ethnic belonging in these areas. This has two consequences. Firstly, we must rely on geographical sampling in areas where we know that there are considerable Inuit or Saami populations. Secondly, that we can only make an ex post delimitation based on the interviews between Inuit and Saami peoples and other people.
The challenge is therefore to set up theoretical and empirical variables regarding the ethnic ties of the individual. According to this we must partly deal with the general scientific debate regarding ethnicity as described in the sociological and anthropological literature and partly we must relate our findings to the populations and areas that are covered by the project.
3.1 The theoretical delimitation
Our first assumption regarding an ethnic group is that membership is not entirely a question of either or being Inuit or Saami. There will, of course, be a group of respondents who can be categorized as not being Inuit or Saami. Regarding a majority of our sample there will however be a considerable proportion of respondents that can be defined as Inuit or Saami to some extent. The point here is that ethnicity is a multidimensional concept and that the ethnic belonging of the individual should be seen as a point on a scale between two extremities. This is why Saami peoples themselves and researchers distinguish between ordinary Saami and "Super-Saami". The latter meaning a person who emotionally and often politically is very engaged or associated with the Saami culture.
Attempting to define an ethnic group as a general concept, one place to begin might be a modified version of Dahlstroems definition of a national minority group from 1971: "A separate group consisting of members with a common descent and language who feel connected to each other by a specific lifestyle and culture".
This definition underlines that an ethnic group is a social system. It means that the members of the group interact to a higher degree with each other that with other people. The main problem with this definition is, however, that it is so demanding that only "Super-Inuit" or "Super-Saami" would qualify as indigenous peoples in our survey. But the definition is interesting because it highlights the criteria which are usually applied in definitions concerning both national minorities and ethnic groups.
These are outlined below:
1. Subjective criteria: Criteria or definitions that emphasize that members of the group feel themselves connected to an ethnic group by emotional ties and who are conscious of this connection.
2. Objective criteria: Definitions which distinguish a specific ethnic group from others by referring to descent, race, religion, language, territory, etc.
3. Deviant culture: Definitions that emphasize the specific culture of a certain ethnic group.
4. Mixed definitions: Definitions that contain two or three of the above-mentioned criteria.
It is our point of view that we need to apply a mixed definition of an ethnic group in our study. If only objective criteria as descent or parents' descent are applied, the result will be a statistical category of persons perhaps with nothing else in common than the applied criterions. Certainly, there will be no way to know whether these people constitute a social system or just a random collection of respondents with only a few things in common.
Using subjective criterions alone is also very risky. Especially because individuals can use their ethnic background instrumentally to gain personal advantages. This has been the case among others in Scandinavia where people have defined themselves as Saami to get land in connection with the land claim negotiations between the Saami and the governments in Finland and Norway. In an attempt to define the concept of an ethnic group, it is therefore crucial to relate to the ontological status of the word.
In sociological and anthropological literature there have been two polarized approaches. The one extreme is "primordialism" according to which the ethnic belonging of the individual is seen as primordial or even given by nature. Opposed to the primordialists are theorists who tend to view ethnicity as a resource to be mobilized by individuals or groups in pursuit of political or economic ends. In this project it is assumed that ethnic ties can be very deep-seated and that the individual might not be aware how these ties have developed. At the same time individuals and groups are often conscious of the fact that they belong to an ethnic group and as seen in Scandinavia can use this back group instrumentally. The point here is that subjective sentiments and ties should only be part of the definition of an ethnic group since the subjective responses regarding ethnic belonging might be unreliable due to the above-mentioned reasons.
It is not appropriate to define an ethnic group based on a common culture alone. It is possible to imagine cross-national common cultures among different religious groups or social classes that cannot be said to be ethnically connected. The members of an ethnic group must have a common descent in some way. Culture defined as "A common pattern of experiences, perceptions, attitudes and knowledge that exist in relation to a social system" should, however, be part of a definition of ethnic groups.
As mentioned above, we think that an appropriate definition of an ethnic group in our context should include objective measures, subjective sentiments/attitudes and culture as defined above. Contrary to Dahlstroems definition we think that it should be stressed that an ethnic group can be more or less integrated and that individuals can feel more or less attached to an ethnic group.
3.2 The empirical delimitation
The point here is to describe how to operationalize the theoretical considerations from the previous section or how to measure the degree of individual ethnicity. It is our aim to construct three empirical measures of ethnicity based on objective criterions, subjective attitudes and culture.
Objective measures
Our primary objective measure will be place of birth of the respondent and his/hers parents. This variable is used in the official statistics in Greenland and Scandinavia as measures of ethnicity.
Subjective measures
Regarding subjective measures it is obvious to ask the respondents which ethnic group they feel attached to. Besides, this measure of salience is needed. It means to ask the respondents how important their ethnic ties are to them.
Culture manifests itself in a variety of ways, and it is rather difficult to measure a common culture in a quantitative survey as this one - but not impossible. In this study culture will be measured by a behavioural cumulative index consisting of a range of questions.
The items in this index must according to Elklit, Noack and Tonsgaard meet four demands:
1. They must represent a broad spectrum of the cultural pattern which the individual is part of.
2. There must be no reason to assume that the different forms of behaviour are determined by factors that are not relevant or connected to the ethnic identity in question.
3. The items must measure actual behaviour and not expected behaviour or attitudes.
4. Fourthly the majority of the respondents must have the opportunity of practising the chosen forms of behaviour.
It is part of the project to operationalize these items, but possible forms of behaviour might be participation in ethnic political and cultural organizations, language, nutritional habits, a specific upbringing of children (e.g. sending children to ethnic schools, working with children) etc.
The overall aim is to obtain three different measures of ethnicity that reflect different aspects of the concept. Using more measures makes it possible to take into consideration the multidimensionality of ethnicity. Besides this it makes it possible to investigate the validity of the separate measures by correlating these. We are of course aware that it might be difficult - if not impossible - to apply the same measures in all the areas in the survey. We hope however by means of these measures to create an instrument to measure the degree of ethnic ties of the individual.
4. The demographic aspects
4.1 The demographic measures
In this section we will focus on fertility and mortality and the connection between these two variables and the possibilities of calculating fertility and mortality rates from data already existing and from data collected during a sample survey.
Reliable information about fertility and mortality is essential to any planning for development. The relationship between births and deaths (in the absence of migration) determines the rate of change in the size of the population, and fertility is a major determinant on the age composition of the population.
In an approach to analyse the living conditions in the Arctic, fertility and mortality rates are good health indicators. The linkage between fertility and mortality consists partly of what has been called a "physiological effect" whereby a child's death truncates breastfeeding, having the women reexposed to the risk of conception and partly of insurance and replacement effects: high levels of mortality are generally associated with high levels of fertility. Mortality affects fertility through a number of mechanisms. Firstly, it affects the number of couples of reproductive age through its general influence on the age and sex structure of the population. Correspondingly, at the individual level, the number of children that a couple is likely to have is influenced by the probability of the full reproductive life of the couple remaining unbroken by the death of one of the spouses. Secondly, infant and child mortality have been hypothesised to affect fertility, through both biological and behavioural mechanisms.
One of the strong determinants of fertility is education, meaning that one of the most important correlates of female education is reduced fertility. The issue of women's education has been discussed in length in the context of child health, and it would be reasonable to assume that it would have a similar positive effect on their own health. The strong linkages between female education and better health are well known. At the macro-level cross-national studies show high correlations between female literacy and life expectancy at birth, higher than any other factor. A review of studies from around the world supports the inverse relationship between infant mortality and mothers' education. Analysing Nigerian data, Caldwell found that mothers' education was a more important determinant of child mortality than mothers' age, place of residence or socioeconomic status, fathers' education or occupation, income or access to health facilities.
The mechanisms whereby women's education results in lower child mortality have been the subject of some speculation. As child health and survival are enhanced by better hygiene, improved nutrition and feeding practices and timely medical intervention, education may improve women's practice of any of these.
Education is said to change the women's status and values away from a traditional lifestyle which includes a large family size, towards a lifestyle which includes a higher status, a career and increasing array of consumer goods. Education also increases a woman's income earning potential and, as a result of this, the opportunity cost of withdrawing from the labour force in order to care for children.
On the basis of the above it can be concluded that in a study of living conditions it willbe essential to include measures of fertility and mortality. Fertility decline is seen as the natural concomitant of economic development, and improvements in mortality rates are likely, in part, to be the result of economic development and service delivery. In a comparative study like this one concerning the Inuit and Saami, populations are experiencing a rapid transition from traditional to modern society, and fertility and mortality are important variables supplying indirect information on the development in health, economic and social conditions. Infant mortality rates are generally considered to be a powerful reflection of underlying disparities in socioeconomic conditions and health care services. Fertility and mortality rates cannot only supply a picture of health conditions in the different areas included in the study of living conditions in the Arctic, but they can also provide an insight into the economic performance of these societies.
4.2 Methods of data collection
4.2.1 The three basic ways to obtain data
There are three basic possibilities of obtaining information on the population:
A census of the population, which provides data on all the people in a specified area at a specified time.
A civil registration system, which seeks to record births and deaths when they occur.
Sample surveys which collect the relevant information from a sample of the population.
The three methods are complementary, each providing some information that is not available from the others. Population censuses
Censuses are generally less effective in providing data on mortality than on fertility. Death is often considered a topic to be avoided, which makes it difficult to gather information.
Censuses also obtain other data that permit the estimation of mortality. Of child mortality by asking the mothers about the survival of their children, of adult mortality by asking people about the survival of their parents or their first spouse.
Of the areas included in the study of living conditions in the Arctic data is partly supplied by censuses in Inuvialuit, Nunavut, Nunavik, Labrador (Census conducted in 1996) and Alaska (Latest census conducted in 1990). Civil registration system
Registration systems differ from censuses and sample surveys in a number of ways. The primary purpose of registration is to establish legally certain facts about an individual, i.e. where and when she was born and where and when she died.
In Greenland, Sweden, Norway, Finland, Chukotka and the Kola Peninsula statistics are based primarily on civil registration systems. The statistics in these countries includes demographic measures for the whole population. The problem using these data in the study of living conditions of indigenous people in the Arctic is primarily that they contain very weak measures of ethnicity. Sample surveys
The sample survey often allows more time for each interview than is possible during a census so that any topic can be explored deeper. In a survey done independently of a census there is more latitude for detailed questioning.
Sample surveys are used in all areas included in the study of living conditions in the Arctic, since a survey gives the possibility of investigating special connections between different variables. Connections that one cannot make via civil registration systems or censuses. A comparison of the three methods
One of the chief advantages of a properly functioning civil registration system is that it can provide timely data on births and deaths at all levels. It should also be able to provide initial summaries of tabulated data relatively soon after the collection of the basic information. The results of a survey can normally be made available more quickly than the results of a census.
A census is a less satisfactory source of information about mortality than other methods (except about child mortality if certain questions are included), whereas civil registration systems give very useful information about fertility and mortality and in the analysis of the determinants of fertility and mortality.
The conditions and the auspices under which information is solicited not only affect the response rate, but also the appropriateness of certain questions and people's willingness to answer them accurately or answer them at all. Both censuses and civil registration systems are usually conducted in accordance with laws that require the respondents to answer fully and to the best of their ability. Sample surveys, on the other hand, are more often conducted on the basis of voluntary participation, which gives the respondent the right to refuse to answer any or all of the questions. Items that are considered sensitive may not be included in a census questionnaire but can be included in the questionnaire of a survey.
In censuses and universal civil registration systems, the results are not subject to sampling errors, because the entire population is included. Results of sample surveys, however, are subject to sample errors, and these may be especially large when detailed cross-classifications are undertaken.
In the case of the study of living conditions in the Arctic, the easiest way to obtain data would be to use the official data from the various places. This is unfortunately not possible since, as already mentioned, most of the areas considered do not have an ethnic variable that is usable in this study. Besides that, the official statistics do not allow for analysis between for example mortality and fertility and other structural variables.
Table 1 Advantages and limitations of censuses, civil registration systems and sample surveys
In Sweden, Norway and Finland there is no official statistics on the Saami population in terms of fertility and mortality. On the Kola peninsula there do exist statistics concerning the indigenous population. Here the problem is the definition of a person of indigenous descent. The criterion for belonging to the indigenous part of the population is very wide: if one fell that one belongs to this group, then one can be registered as such. In Greenland the vital registration system registers ethnicity, but on the basis of place of birth. There are two different categories: born in Greenland and born outside Greenland, as already mentioned in section 3.
In the light of the above it seems that the only choice left is to conduct a survey in the areas in question. This does not mean, however, that one cannot use the data already available. These data can serve as a standard of reference for the results that will be obtained in the survey.
4.3 Fertility
The main fertility and fertility related items generally collected through household surveys are as follows:
Basic items included in most population censuses and surveys are the number of children ever born alive (and number still living) by women in the population. Classified by the mother's age, these items provide information on "cumulative" or lifetime fertility and on "traditional" levels of fertility and family size on the basis of information reported by older women. Such information may also be used to estimate current levels of fertility. The period total fertility rate is a measure of completed family size. Since TFR combines the current fertility experience of females 15 - 49 years, it can be interpreted as a hypothetical measure of completed cohort fertility.
In a household survey, data on current fertility may be obtained more directly from retrospective questions on births occurring in a specified period of time (during the last 12 or 24 months from the date of the last birth).
In many societies childbearing takes place largely within the context of marriage. Apart from data on marital status, a survey may also collect data regarding age at first marriage, duration of marriage, or even full marriage histories. In the case of Greenland, childbearing mostly takes place outside marriage and if this is the case in the other areas, there will be little meaning using resources on collecting detailed marital data. It can therefore be a good idea to add a definition of couples living together without being married as a supplement to the data on marital status.
4.3.1 Definitions
In the following issues concerning concepts and definitions of demographic characteristics and collection of these statistics in a household survey will be discussed. Items for fertility estimations to be discussed are age at marriage and duration of marriage, children ever born alive and children still living, and live births within 12 months preceding the survey. Age at marriage and duration of marriage
Age at marriage is the age of the woman, in completed years, at the time when the marriage took place. Duration of marriage which is the interval between the data of marriage and the reference date of the survey or the date of the dissolution of the marriage prior to the survey is expressed in completed years. The questions regarding age at marriage and duration of marriage are normally related to women in their first marriage, and those widowed, divorced or separated who have only had one marriage. Children ever born alive and children living
Information on the number of children born alive and the number of children still living is collected from women above a specified age. The number of children born alive should include all children born alive during the lifetime of the women up to the survey date, whether born during the present or a prior marriage. The question should be directed to all women above the specified certain age (ranging from 12 to 18 years) regardless of marital status. It has not yet been decided what this age should be in the study of living conditions in the Arctic, but it will properly be 15 years. Children still living include all the children who have been born alive and who are still alive. It should therefore exclude foetal deaths and step or adopted children. Live births within 12 months preceding the survey
Information on live births occurring to women within the last 12 months preceding the survey, may be used to measure current fertility. However, due to errors in dating and omissions due to women who for example are not reporting children born 11 months ago, this question alone cannot produce a reliable estimate of current fertility. The question should be asked directly to women above the specified age. It is very common in collecting this information for births occurring more than a year ago to be included or births occurring almost a year ago to be excluded.
Another problem, which often arises, is that the proportion of women who have given birth during the last 12 months before the survey is relatively small. The information on births occurring during 12 months preceding the survey can be approximated through the question of the last live birth, since in this case an answer must be given for every woman who has at least had one live birth in her lifetime. However, the last question will not yield the number of children born in the last 12 months preceding the survey, but only the number of women who have had their last live birth during the last 12 months.
4.3.2 Estimates of fertility based on information about children ever born General description of methods of the Brass type
The total number of children ever born by a group of women of a given age is a record of their total childbearing experience from the beginning of their reproductive period to their current age. The average number of children ever born is therefore a measure of the fertility experience of the cohort of women, though it is a measure of the level of fertility only, containing no information on its timing. If it is assumed that the fertility experience of those women who die is the same up to the age of death as that of those who survive, the average number of children born provides a mortality-free measure ofcohort fertility.
The availability of information about lifetime fertility, from a survey question about number of children ever born, and current fertility, from a question on births in the past year or date of the most recent birth makes possible a consistency check, whereby current fertility rates can be cumulated and compared with average parity. Such a comparison clearly uses both cohort rates and period rates, but it is valuable even if the two are not expected to be consistent because of changing fertility.
Information on children ever born is often distorted with omissions. One reason is that the woman does not report children who have died a very short time after being born, but this omission is most frequent for older women. The report of younger women, up to 30 or 35, may be fairly reliable. Information on current fertility may be distorted by a misperception of the length of the reference period, so that the reported births correspond to an ill-defined period whose average length may be shorter or longer than a year.
Cumulated current fertility may be compared with the reported lifetime fertility of women younger than 30 or 35 in order to obtain an adjustment factor for the level of the current fertility rates, which adjusted for level, provide a better estimate of current fertility. In order for this adjustment to be valid, it must be assumed that the fertility of the younger women has not changed appreciably. Furthermore, when dealing with data classified by five-year age groups, cumulated current fertility rates provide an estimate of the average number of children ever born by women who have reached the end of each age group, whereas parity data provide an estimate of the average number born by women whose ages vary over the range of the age group. Therefore a process of interpolation is required to ensure that the figures cover a comparable age range.
The essence of the Brass fertility estimation procedure is the adjustment of the age pattern of fertility derived from information on recent births by the level of fertility implied by the average parity of women in age groups 20-24, 25-29 and perhaps 30-34. Between the age 15 and 30 years most individuals form relationships and have their first children. These individuals often migrate or set up new households. The significance of these changes is made even more crucial by the fact that the Brass technique as mentioned rely on the reported experiences of women in these ages. Migration and formation of new households can be responsible for some underenumeration. It has for example been pointed out that in Cameroon a young woman who is engaged to be married, but still living with her parents, might not be counted in neither her parents home, nor in the home of her coming husband since she does not yet belong any of the places.
Reported fertility rates are used to estimate the average cumulated fertility, F, that women in each age interval would have if they had been subject throughout their lives to the reported rates. Two problems arise in obtaining a value of F that is comparable to the average parity, P, reported by women in each age group. First, because fertility data are ordinarily tabulated by five-year age groups, cumulating the reported age-specific fertility rates, and multiplying by five yields estimates of the cumulated fertility that women experience those rates would achieve by end of each five-year age group (that is at exact ages 20, 25, 30 etc. where conventional age groups are used).
These estimates are not comparable with the average parities calculated from data on children ever born, because the latter values represent the mixed experience of women of different exact ages. Some procedures are required for estimating the average cumulated fertility or parity within each age group from knowledge of the values that the cumulated fertility schedule takes at the endpoint of the age groups considered. Secondly, when the current fertility schedule is obtained from a survey question on births during the 12 months preceding the survey, the births are generally tabulated by the mother's age at the time of the survey, and not at the time of the birth.
If one assumes that births in a given year are uniformly distributed over the year, the women who had a birth 12 months preceding the survey were in average 6 months younger at the time of the birth than at the time of the survey. Therefore the age-specific fertility rates that can be calculated from data on children ever born during the year preceding the survey classified by the age of the mother at the time of the survey correspond to unorthodox age intervals whose limits are (14,5, 19,5), (19,5, 24,5), ....... , (44,5, 49,5) rather than to the usual intervals with endpoints (15, 20), (20,25), ...... , (45, 50).
An interpolation procedure based on model fertility schedules has been devised to allow the estimation of the average cumulated fertility, F, for the usual five-year age groups of women from the cumulated fertility schedules. A similar procedure that takes into account the problem of age groups displaced by six months and produces the desired estimates of F has also been developed. The Brass method and its rationale
The Brass method seeks to adjust the level of observed age-specific fertility rates, which are assumed to represent the true pattern of fertility, to agree with the level of fertility indicated by the average parities of women in the age groups younger than ages 30 or 35, which are assumed to be accurate. Measures of average cumulated fertility, F, comparable to reported average parities, P, are obtained from period fertility rates by cumulation and interpolation. Ratios of the average parities, P, to the estimated average cumulated fertility, F, are calculated age group by age group, and an average of the ratios obtained for younger women is used as an adjustment factor by which all the observed period fertility rates are multiplied. Data required
The number of children ever born classified by five-year age group of the mother.
The number of children born during the year preceding the survey classified by five-year age group of the mother.
The total number of women in each five-year age group (irrespective of marital status)
The total population if the birth rate is to be estimated. Possible questions in the survey
.Have you ever given birth ?
.Do you have any sons or daughters to whom you have given birth and who are now living with you?
.How many sons live with you ? and how many daughters live with you ?
.Have you any sons or daughters to whom you have given birth and who are alive but do not live with you ?
.How many sons are alive but do not live with you ? And how many daughters are alive but do not live with you ?
.Have you ever given birth to a boy or a girl who later died ?
.In all how many boys have died ? And how many girls have died ?
4.4 Mortality
Data on mortality collected through surveys have included the following:
In many developing countries the only source of information on infant and child mortality have been simple questions on lifetime fertility (children ever born) and proportions of children dead for each age group of mothers. These probabilities have been converted into estimates of lifetime probabilities of dying by specified ages, such as 1, 2 and 5 years, using indirect estimation techniques.
Procedures have also been developed to obtain indirect estimations of adult mortality on the basis of information on the survivorship of parents or spouse. The indirect questions are considered rather simple in principle and their use depends on certain assumptions regarding demographic patterns and trends.
For direct estimation of mortality it is necessary to record deaths (classified by age at death, sex and the like during a specified period of time, such as 12 or 24 months). In a single-round survey this takes the form of retrospective questioning about deaths in the household. For infant and child mortality estimations, the information may be collected in the form of retrospective birth histories of mothers, in which data of birth, sex and survivorship status of all children alive are recorded. In a less elaborate form, such information may be obtained only for the last or most recent birth to estimate current levels of infant mortality.
4.4.1 Definitions
The items regarding estimation of mortality which will be discussed in the following section are deaths within 12 or 24 months preceding the survey and orphanhood and widowhood. The data on children ever born and children surviving discussed above in connection with the estimation of fertility are also used to estimate infant and child mortality. Deaths within 12 nor 24 months preceding the survey
There are two types of questions which are usually used: (a) deaths of infants born within 12 months preceding the survey, for the measurement of infant mortality, and (b) deaths of members of the household occurring within 24 months preceding the survey, which is primarily used to derive adult mortality. Experience shows that neither of these questions have given satisfactory data, and as in case of births the results also suffer from dating errors and omissions errors, as well as low incidence in the sample. The first question is normally asked in combination with births during 12 months preceding the survey. The second question will produce useful results if it can be assumed that the completeness of the reporting of deaths is the same for all ages. If so, an analytical technique has been developed to adjust the observed death rates to obtain a better estimate of true mortality conditions. In both cases the data collected should include sex, age and month of occurrence. Orphanhood and widowhood
Information on maternal and paternal orphanhood, which is concerned with the survival of the first spouse of the respondent, can often be collected through household surveys. This information will be useful in the estimation of adult mortality. It should be mentioned that the term "orphanhood" is used in a rather unusual sense here, since it is specific to a parent's sex. An orphaned respondent is one whose mother (father) has died, regardless of the farther's (mother's) survival. Estimation of adult mortality through household surveys has not produced satisfactory results, but if these data can be collected with sufficient accuracy, they can be used as an alternative method for obtaining adult mortality estimates. It should be made clear that the mother and father in question are the biological parents of the respondent. In a society where adoptions are common it often happens that the respondents do not even know that they are adopted. In addition to this information, the estimation technique also requires the age at marriage of the respondent's first spouse.
4.4.2 Estimates of infant and child mortality General description of the method
The births to a group of women follow some distribution over time, and the time since birth is the length of exposure to the risk of dying of each person. The proportion dead of children ever born to a group of women will therefore depend upon the distribution of the children by length of exposure to the risk of dying (the distribution of in time of the births) and upon the mortality risks themselves. By allowing for the effects of the distribution of the births in time, the proportion of dead children can be converted into a mortality measure expressing the average experience.
Brass developed a procedure to convert the proportion of dead children ever born to women in successive five-year age groups into estimates of the probability of dying between birth and an exact age. Brass found that the relation between the proportion of children dead and a life-table mortality measure is primarily influenced by the age pattern of fertility, because it is this pattern that determines the distribution of the children of a group of women by length of exposure to the risk of dying. He developed a set of multipliers to convert observed values of D(i), the proportion of children dead, into estimates of q(x), the multipliers being selected according to the value of P(i)/P(i+1) - a good indicator of fertility conditions in the younger ages - where P(i) is the average parity or average number of children ever born reported by women in age group i.
An important assumption made in the development of this method is that the risk of a child dying is a function only of the age of the child and not of other factors, such as mother's age or the child's birth order. In practice it appears that children born by young mothers experience mortality risks well above average. For this reason the estimate of the infant mortality rate, q(1) the probability of dying before the age of 1, that can be derived from reports of women in the age group 15-19, are generally excluded. This is also the case because the number of children born and dead is usually small.
Trying to increase the flexibility of Brass' original method, Sullivan computed another set of multipliers and Trussel estimated a third set of multipliers by the same means but using data generated from the model fertility schedules developed by Coale and Trussel. The general theory on which these methods are based is essentially the same, but they arrive at somewhat different multipliers because the data bases is different in each case. Since the Sullivan variant has no obvious advantages over that proposed by Trussel, whereas the latter is based on a wider range of cases, the Trussel procedure is described here.
It is important to be aware that the original methods of estimation are based on the assumption that fertility and childhood mortality have remained constant in the recent past. If fertility has been changing, the ratios of average parities obtained from a cross-sectional survey will not replicate accurately the experience of any cohort of women and will not provide a good index of any distribution in time of the births to the women of each age group.
The conversion of the proportion of children dead into conventional mortality measures also requires the use of mortality models, and the age pattern of mortality assumed has some impact on the q(0) estimate obtained. Versions of Sullivan's and Trussel's procedures are available for each of the four regional model life tables of Coale and Demeny. The choice of model can affect the estimates of q(0) from reports of women under 30 by nearly 10 percent, so it is helpful to have some evidence on which to base such a choice. If no suitable evidence exists, the assumption of a "West" pattern introduces the least potential error, and the estimates of q(5) and q(10) from reports of women aged 34-39 and 35-39, respectively, are little affected by the mortality pattern, though extrapolation from these values either to an infant mortality rate or to mortality in later years is misleading. When in doubt the calculations can be repeated for all four mortality patterns, and this is under consideration in the study of living conditions in the Arctic.
Mortality change is generally taken into account by assuming that child mortality has been changing in a regular way over time. Model calculations have shown that under such circumstances a child mortality estimate derived from a proportion of children dead closely approximates the period life table for a particular point in time prior to the collection of data, regardless of the pace of mortality change. Data required
The number of children ever born, classified by sex and by five-year age group of the mother.
The number of children surviving (or the number dead) classified by sex and by five-year age group of the mother.
The total number of women (irrespective of marital status) classified by five-year age group of the mother. Possible questions in the survey
.Have you ever given birth ?
.Do you have any sons or daughters to whom you have given birth and who are now living with you?
.How many sons live with you ? and how many daughters live with you ?
.Have you any sons or daughters to whom you have given birth and who are alive but do not live with you ?
.How many sons are alive but do not live with you ? And how many daughters are alive but do not live with you ?
.Have you ever given birth to a boy or a girl who later died ?
.In all how many boys have died ? And how many girls have died ?
4.4.3 Estimates of adult mortality
This section describes how to obtain estimates of adult mortality from information concerning the survival of parents and spouses. In both cases a particular target person is known to have been alive at a specific time related to a specific past event and some information is available about both the length of exposure to the risk of dying and about the age at which the exposure began. General description of the method
Information on the survival of parents or spouse is an indicator only of adult mortality,since the exposure to risk of the target person begins in adulthood, at the birth, conception or marriage of the respondent. Such data should only be used to estimate survivorship probabilities between one adult age and another. However, if an estimate of the level of child mortality is available, and some assumptions can be made about the form of the relationship between child and adult mortality in the population under study, the information on child mortality can be combined with an indicator of adult mortality to estimate an unconditional survivorship probability, that is the probability of surviving from birth to some adult age. In choosing between these two approaches, it should be remembered that the first approach is less dependent on assumed models, but it is also more difficult to incorporate into a full life table, than is the latter approach, which in turn is determined, to a considerable extent, by the estimates of child mortality used in its calculation. In this paper we will only focus on the first approach.
Estimates of adult mortality derived from information on the survival of close relatives represent averages of the mortality experienced over the period during which the relatives were exposed to the risk of dying. If mortality has changed in a regular way, each estimate has a time reference, i.e. there is a time before the survey with a period life table having the survivorship probability that has been estimated. The number of years, t, before the survey that defines the period to which this life table refers will depend mainly on the average exposure to risk of the target persons and on orphanhood or widowhood status of respondents.
The method described in this paper deals with the estimation of adult mortality from information about the orphanhood or widowhood status of respondents. This method may be called "conditional" because the estimated parameters are conditional probabilities of survival that do not, by themselves, define a complete life table. To derive life table values from them, more information is needed about child mortality. Basis of the method
Some general features about the method should be pointed out. First, the estimated probabilities of survivorship do not refer to the entire population, since they reflect only the mortality experience of parents with surviving children. Secondly, survivorship estimates based on reports by young residents, and thus corresponding to small values of n (under 20), tend to be affected by misreporting of orphanhood status: young orphaned children are often adopted by relatives who report them as their own children. This phenomenon inflates the proportion of young respondents having a surviving parent and biases upwards the estimated survivorship probabilities of younger adults. Lastly, the estimated probabilities of survival do not, strictly speaking, refer to specific time periods, since they represent average measures over the somewhat ill-defined intervals of exposure to the risk of dying of the target population. In cases where mortality has remained essentially constant, problems in the interpretation of the estimates obtained do not arise since, in the absence of data errors, they should all imply the same, unchanging mortality level. However, since mortality has not remained constant in the recent past in most countries, the interpretation of survivorship probabilities derived from orphanhood data is not always straightforward.
One starts the computation procedure with calculation of mean age at maternity (paternity). The births used to calculate the mean age, M, are those reported as occurring during the year preceding the survey and are tabulated by age of the mother at the time of the survey. The true age groups of the mothers will on average be six months younger than stated, so that six months should be subtracted from the midpoint of each age group when calculating M.
The estimation of M for males is one of the additional problems associated with the estimation of adult mortality from the proportions of respondents with a surviving father. Fertility questions are generally not posed to males. It means that the information from which the female M is estimated is usually not available for fathers. Births during the year preceding a survey are sometimes tabulated by age of the husband, but this tabulation is generally limited to those cases in which the mother and her husband are enumerated in the same household. Calculating the male M from such a tabulation often biases its value upwards because young fathers are more likely to be temporarily absent.
A more robust procedure for estimating M for males is by adjusting the female M by using information on marital status. The median age of the currently married population can be calculated by sex, and the difference between the male and female medians can be added to the previously calculated female M to obtain an estimate of the male M. Medians are used to reduce the influence of older, and for this purpose, irrelevant couples. Data required
The proportion of respondents with a surviving mother (father) in each five-year age group from n to n+4. The set of proportions can be calculated when any two of the following items are available:
(a) The number of respondents with the mother (father) alive classified by five-year age group.
(b) The number of respondents with the mother (father) dead classified by five-year age group.
(c) The total number of respondents whose mother's (father's) survival status is known classified by five-year age group.
The number of births in a given year classified by five-year age group of the mother (father). This information is needed to calculate the mean age of mothers (fathers) at the birth of their children in the population being studied. Possible questions in the survey
.Is your father alive ?
.Is your mother alive ?
.Is your first spouse still alive ?
5. Perspectives
In the previous sections we have outlined a sketch of two crucial methodological aspects of this project. The above discussions are preliminary and will be debated further during 1998 in connection with the completion of the research design of this project.