470 likes | 476 Views
European Commission – Eurostat. Second meeting of the Technical Coordination Group for the Censuses in South East Europe. Dissemination of statistics on ethnic/religious/language groups Donatella Zindato Istat, Central Directorate for General Censuses
E N D
European Commission – Eurostat. Second meeting of the Technical Coordination Group for the Censuses in South East Europe Dissemination of statistics on ethnic/religious/language groups Donatella Zindato Istat, Central Directorate for General Censuses Head of Population and Housing Census Operational Planning Unit Skopje, State Statistical Office – 24-25 June 2010
General recommendations on ethno-cultural topics Multi-ethnic countries with long-established minorities and/or recently arrived immigrant populations may wish to collect information on the ethnic composition of the population or of certain subgroups of the population. The data are relevant for the understanding of the cultural diversity of the population, the position of ethnic groups in society as well as for the definition and monitoring of anti-discrimination policies (par. 420 ). Strong subjective dimension and sensitivity of ethno-cultural topics. All possible measures to guarantee that the free self-declaration of the respondent is recorded and preserved during all enumeration phases (data collection, coding, processing, editing and dissemination) data should reflect the reality of the country and its population (the ‘reality’ is what is reported by respondents)
Ethno-cultural topics: data collection • Respondents should be informed about their right not to reply. • Interviewers should always ask the question without suggesting or influencing the answer assuming that they know it. • Information on one topic cannot be derived from information on another one (for instance, religion cannot be derived on the basis of ethnicity).
Ethnicity • Concept difficult to measure • Subjective • Multi-dimensional • Fluid (it tends to change over time) Membership of an ethnic group is something that is subjectively meaningful to the person concerned Two main factors appear to be determining the ethnic group recorded for each respondent • its own choice of how he/she views his/her own ethnicity • the question format
Methodological and operational issues (1) • The accuracy of responses (level of detail and consistency) may be affected by • the structure of the question itself • the selection of answer categories or examples • ⇨ • information on ethnicity has to be acquired through self-declaration • respondents have to be given the option of indicating multiple ethnic affiliations • the selection of answer categories or examples has to be carefully planned in order to achieve an accurate reporting of ethnicity data • when the answer categories or the examples included in a question change, comparability of data from one census to another might be affected
Methodological and operational issues (2) • No objective and rigid classification • No internationally recognized classification • Need of ethnic classification standards that are simple and stable, but also sufficiently flexible to cope with the complex and changing nature of ethnicity
When it comes to dissemination… (1) • The method and the format of the question used to measure ethnicity can influence the choice that respondents make regarding their ethnic background/identification • No objective and rigid classification • Countries should document the basic criteria and classification procedures for ethnicity and inform the data users about the scientific and socio-political concepts on which they are based
When it comes to dissemination… (2) • The basic criteria used to measure the concept have to be clearly explained to data users • concept definition • question format • list of examples provided and criteria for choice • additional instruction on how to complete the question (if any) • criteria used for determining the ethnicity of children from mixed couples • Output classifications should be extensive and discussed with ethnic groups themselves • classifications of ethnic groups at the finest level should be comprehensive and include ethnic groups, self-perceived groups, regional and local groups as well as groups that are usually not considered to be ethnic groups • ethnic group data should be presented in as much detail as possible
Dissemination of multiple ethnic affiliations • Results should be presented also for multiple ethnic affiliations and combinations of ethnic affiliations, as well as for the categories ‘no ethnicity’ and ‘not declared/prefer not to answer’.
Ethnicity: aggregated classifications • aggregations should be documented and fully transparent • broad headings should be avoided • broad headings such as ‘Asian’ or ‘Asian British’ will mask important distinctions, such as those between the Indian, Bangladeshi and Pakistani ethnic groups ⇨ need to distinguish between each of these groups wherever possible • in case that data for some ethnic groups might not be published separately due to confidentiality issues (disclosure risks), aggregations should fit the data • to aggregate data by combining different ethnic groups should only be done if the two groups show similar patterns on the outcome variable of interest • rather than combining categories inappropriately, it is better to show a category in a table while indicating that data for that category have been omitted for confidentiality reasons
On the other hand… • Some respondents may choose to provide very specific ethnic origins in the census, while others may choose to give more general responses. This means that two respondents with the same ethnic ancestry could have different response patterns and thus could be counted as having different ethnic origins. • for example, one respondent may report 'Black' while another, similar respondent, may report 'Ghanaian' or 'African'. As a result, counts for certain origins may seem lower than initially expected. Users who wish to obtain broader response counts should be advised to combine data for two or more ethnic origins together.
Data quality documentation • For the ethnic origin variable, the assessment of data quality was based on a review of response rates, imputation rates, and a comparison of 2006 counts to those obtained in previous censuses. The overall quality of the ethnic origin variable for the 2006 Census is high. • The 2006 non-response rate was 5.5%, compared to 5.6% in 2001. • Two types of imputation were applied to the ethnic origin data, namely deterministic imputation and donor imputation. • Deterministic imputation is the process by which a unique value is assigned to a missing or invalid response through either relationships among personal characteristics, or, in the case of children with no responses to the ethnic origin question, by using the ethnic origin(s) of their parent(s) to fill in the missing data. • Donor imputation is performed by ... [Statistics Canada, Ethnic Origin Reference Guide, 2006 Census]
Historical comparability (1) Ethnic groups however defined or measured will tend to change over time a person may record themselves as belonging to one ethnic group at one time and to another one on a subsequent occasion Such changes depend upon • genuine change of ethnic identification (due to social and political attitudes and developments) • respondent error in completing the form • proxy responses (the ethnic group of one individual is entered by another member of the household) • question change • erroneous imputation (or differences in the treatment of non-response)
Reporting of Canadian as an ethnic origin • Changes to the ethnic origin question have affected the number of people reporting Canadian as part of their ethnic heritage. • When Canadian was not listed as an example on the 1991 Census questionnaire, only 2.8% reported Canadian as their only ethnic origin and 1.0% reported it in combination with other origins. • Canadian was added as an example on the English questionnaire (and Canadien on the French questionnaire) because this was the fifth most frequently reported origin in the 1991 Census • about 8.8 million people reported Canadian, either alone or in combination with other origins, in 1996 • 11.7 million, or 39.4% of the population, reported Canadian as their ethnic ancestry in 2001 • 10.1 million people reported Canadian (as their only ethnic origin or in combination with other origins) in 2006
Historical comparability (2) In order to increase the reliability of comparisons between censuses (when changes are made to the questions) ethnic group categories might be combined • classifications in which more groups are combined offer greater stability, but the number of recognisable ethnic groups which can be analysed is smaller • with each re-classification, individual ethnic groups become invisible as they are merged with other groups • the new composite ethnic categories cannot be viewed as cohesive ethnic groups. [An important element of any ethnic group is that members share a number of characteristics, such as e.g. language, country of origin, religion or culture. In addition, for a collection of people to be considered an ethnic group it is important that they recognise themselves to be a distinct group. These criteria are not always met with the new composite ethnic groups.]
Historical comparability (3) • In the UK Census, the changes to the ethnic group question between the 1991 and the 2001 Census provided respondents with the opportunity to identify as belonging to different ethnic groups in 1991 and 2001. • For example, people who selected the categories ‘Other Black’ or ‘Other Ethnic Group’ or wrote in a ‘mixed’ ethnic description in 1991 were able to choose from four new ‘Mixed’ categories in 2001. • As a result, there were wide variations in the consistency of answers. Some ethnic groups, particularly the White group, were very stable, while other ethnic groups were far less stable. • In most cases where a change in ethnic identification occurred, the change was to a new category that was not available in 1991. • Other changes were more anomalous (due to a genuine change in subjective identity or to unreliability of previous data)
Historical comparability (4) • The LFS survey conducts a series of repeated interviews with respondents over the period of a year. • Up until 2000, LFS respondents had been asked the 1991 Census ethnic group question. In 2001, the new 2001 Census question was introduced on the LFS. • Some respondents in the 2000/2001 LFS were therefore asked both the 1991 and 2001 Census versions of the ethnic group questions.
Historical comparability (5) In order to increase the reliability of 1991 and 2001 Census comparisons ONS has proposed five re-classifications. Users are advised that • consideration about which classification to adopt should take account of the balance between reliability and ethnic group detail • the benefits and disadvantages of each re-classification are considered • the decision about which classification to use will depend on the specific analysis [Karin Bosveld, Helen Connolly, Michael S. Rendall, 2006, A guide to comparing 1991 and 2001 Census ethnic group data. Providing users with advice on using the 1991 and 2001 Census ethnic group questions to examine change in ethnic populations]
Just 7 per cent of those classified to ‘Other Ethnic Group’ in 1991 retained the same label in 2001. With the new classification this increases to 34 per cent.
The eight-category classification combines respondents identifying as ‘Other Black’, ‘Other Asian’ or ‘Other Ethnic Group’ in either census (Table 11). Respondents identifying with any of the ‘Mixed’ ethnic groups in 2001 are also included in the new ‘Other’ category.
Although the new ‘Other’ category can be compared with somewhat greater confidence (the percentage of those who retain the same label increases to 63), the heterogeneity of the group has increased: • the new ‘Other’ category includes people with a wide variety of ethnic characteristics they would not recognise themselves as sharing a common ethnic identity • they cannot be viewed as a cohesive ethnic group
Religion: data dissemination • Census documentation should clearly explain the chosen concepts and definitions. • Criteria used for determining the religion of children of mixed couples should be clearly indicated. • Concerning the dissemination of census results on religion, particular care should be taken, given the sensitivity of the topic. • Classifications should be comprehensive and include on the finest level also subsets of religions and churches and groups that may not generally considered to be religions.
Religion: concepts and definitions • Statistics Canada, 2006 Census • Respondents were instructed to report a specific denomination or group, even if they were not practising members of their group. • For infants or children, respondents were instructed to report the denomination or group in which they will be raised. In most cases, this would be the religion of their parents (or guardians). • Persons who had no connection or affiliation with any religious group or denomination were instructed to mark the circle "No religion". • However, if respondents considered terms such as "atheist" or "agnostic" to be applicable to them, they were instructed to specify them in the write-in area of the question.
Religion: concepts and definitions • In Northern Ireland censuses traditionally include a question on religion. In 2001 the ‘current religion’ question was extended to include a supplementary question on ‘religion brought up in’, to be answered only by those without a current religion. • A person’s religion has been obtained solely from the response to the current religion part and the output classification includes a ‘No religion and religion not stated category’ (missing answers are not imputed). • Responses from the supplementary question (religion brought up in) have been combined with responses from the current religion part to derive a new output variable, Community Background [responses from the supplementary question are not available separately] • the Community Background variable records a person’s current religion, if any, or the religion brought up in for those people who do not regard themselves as currently belonging to any religion. It includes a ‘None’ category.
Religion: classifications • In Northern Ireland the classification used in standard output has 8 categories • Catholic • Presbyterian Church in Ireland • Church of Ireland • Methodist Church in Ireland • Other Christian (including Christian related) • Other religions and philosophies • No religion or religion not stated • but a complete listing of all religious groups (with ten or more adherents) is given in a standard table.
ONS, Census 2001 Definitions - Output classifications part 2 http://www.statistics.gov.uk/statbase/Product.asp?vlnk=12951
Source: ONS, Census 2001 Definitions - Output classifications part 2 [additional codes available only in in Northern Ireland]
How to present and discuss religion data Since there are several dimensions to religion, talking about ‘religion’ or specific religions without qualifications may be confusing and could be misinterpreted by data users. ‘What is your religion?’ = measure of affiliation i.e. identification with a religion irrespective of actual practice or belief [no measure of belief or belonging] • it is recommended to use the term ‘religious affiliation’ rather than ‘religion’ when referring to census data (e.g. ‘the 2001 Census revealed that 77% of people in England and Wales had a religious affiliation’) • the term faith should not be used as an alternative to religious affiliation, e.g. it would be misleading to say that 77% of people in England and Wales have a faith) • it may be misleading to state ‘Chinese people are the least religious; in 2001 more than half said they had no religion’. More appropriately it should be said that Chinese people are the least likely to have a religious affiliation’. [ONS, Guidance on presenting and discussing religion data Addendum to Ethnic Group Statistics: A guide for the collection and classification of ethnicity data]
Language: data dissemination • Census documentation and reports should clearly explain chosen concepts and definitions and document the classification procedures for languages. • The criterion for determining language for children not yet able to speak should be clearly indicated. • Classifications should be comprehensive and include on the finest level all language groups, regional dialects, sign languages and so on.
Language: data collection • Statistics Canada, 2006 Census • mother tongue • knowledge of official languages • knowledge of non-official languages • home language question • language of work question • To assist people whose first language was neither English nor French, the census questions were translated into 62 other languages, including 18 Aboriginal languages. • 2006 Census data on languages were also obtained for persons living in Indian reserves, Indian settlements and in remote areas, but the examples, where provided for write-in responses, were for Aboriginal languages specifically.
Language: data collection • Mother tongue is the only language question asked of the entire population. • Defined as first language learned at home in childhood and still understood by the individual at the time of the census. • Only the two official languages, English and French, appear on the questionnaire. Other languages can be written in the space provided. • To facilitate the task of respondents, an instruction reads as follows: ‘If this person no longer understands the first language learned, indicate the second language learned.’
On the French version of all census forms, for all questions in the language module where there is a choice of response available, the order in which the choices appear is modified in order to give precedence to the category 'French'.
Language: data collection • Furthermore, the following instructions were provided to respondents in the 2006 Census Guide: • For a person who learned two languages at the same time in early childhood, report the language this person spoke most often at home before starting school. • Report two languages only if they were used equally often and are still understood by this person. • For a child who has not yet learned to speak: • report the language spoken most often to this child at home. • report two languages only if both languages are spoken equally often so that the child learns both languages at the same time.
Language: data quality documentation (1) Statistics Canada, Languages Reference Guide, 2006 Census The overall quality of 2006 Census data for the language variables is considered to be good. • This assessment was based on a review of response rates, imputation rates, a comparison of 2006 counts to those obtained in previous censuses, and a comparison of findings compared to the data from the Landed Immigrant Data System (LIDS) database of Citizenship and Immigration Canada (CIC). • 2006 non-response rates ranged from a high of 3.5% for the question on knowledge of non-official languages to a low of 2.1% for the question on mother tongue (20% sample). • Imputation rates ranged from a high of 1.9% for the question on knowledge of non-official languages to a low of 0.9% for the question on home language.
Language: data quality documentation (2) • 2006 Census counts for the specific languages included in each language variable are generally at expected levels … However, data users should be aware of the following issues: • According to studies on data certification, the 2006 Census statistics on knowledge of official languages could underestimate the category 'English and French' and overestimate the category 'French only,' particularly for the francophone population and, therefore, for the whole population in general. The proportion of Francophones who reported being bilingual was slightly lower in 2006 than in 2001... • When comparing the 2006 Census results to those of the 2001 Census, it appears that there is some overestimation of persons reporting Siouan languages (Dakota/Sioux) in British Columbia and, as a result, also at the Canada level. Although it affects a relatively small population, it is best to apply caution when analysing the census data for Siouan languages (Dakota/Sioux) in these geographies.
Language: historical comparability Over time, the format and wording of census questions is updated to better reflect information about respondents understanding of those questions or to respond to data capture requirements. As a result, data users must be aware of these differences when comparing data over time. There were no changes to the wording, instructions or response categories in any of the language questions asked on the census questionnaire between 2001 and 2006 except for the fact that the number of write-in spaces provided for the knowledge of non-official languages question (Question 14) was reduced to two lines in 2006 from three lines in 2001. 2006 Census data showed that this change in questionnaire design did have an impact on the multiple responses patterns.
Language: historical comparability • Compared to 2001, there was a drop in both the count and the percentage of respondents reporting knowledge of three or more non-official languages (from 0.6% in 2001 to 0.1% in 2006), while there was an increase in the knowledge of two non-official languages in 2006 (from 2.5% in 2001 to 3.2% in 2006). • The combined percentage of these two categories remained fairly stable between 2006 and 2001 censuses (the distribution of knowing two or more non-official languages was 3.3% in 2006 and 3.0% in 2001). • In 2006, a number of new language categories were released separately. In order to compare the data between categories in 2006 and 2001 censuses, data users should consult the following additional information for comparing 2006 and 2001 language categories. • the 2001 category of Serbo-Croatian is equivalent to the sum of the 2006 categories of Serbo-Croatian and Bosnian… • the 2001 category of Turkic languages …
Lessons learnt • Give users as much information as possible on the concepts, definitions and classifications which lie behind the data produced • Assess data quality and provide users with relevant information (data quality assessment and documentation). • Be aware of any possible differences in the questions (format, wording, instructions, response categories, examples chosen, length of the write-in spaces) that might have an impact on differences over time and provide users with the necessary information for comparing data over time.
Lessons learnt • Give users as much information as possible on the concepts, definitions and classifications which lie behind the data produced • Assess data quality and provide users with relevant information (data quality assessment and documentation). • Be aware of any possible differences in the questions (format, wording, instructions, response categories, examples chosen, length of the write-in spaces) that might have an impact on differences over time and provide users with the necessary information for comparing data over time.
Lessons learnt • Results should be presented also for multiple ethnic affiliations and combinations of ethnic affiliations, as well as for the categories ‘no ethnicity’ and ‘not declared/prefer not to answer’. • No imputation of missing answers if the question is voluntary. • Information on one topic cannot be derived from information on another one (for instance, religion cannot be derived on the basis of ethnicity) no consistency controls and no imputation on the basis of relationships among personal characteristics
USA TODAY Diversity Index, based on the percentage of each race/ethnicity to calculate the chance that any two people are from different groups. The range is from 0 (no diversity) to 100. The 2009 DI is 52 (the chance of two people being ethnically different is slightly more than half (it was 34 in 1980).
Main references • Aspinall Peter J, Operationalising the collection of ethnicity data in studies of the sociology of health and illness, Sociology of health and illness, 2000, 23, 6, 829-862. • Bosveld K., Connolly H., Rendall M.S., 2006, A guide to comparing 1991 and 2001 Census ethnic group data. Providing users with advice on using the 1991 and 2001 Census ethnic group questions to examine change in ethnic populations, ONS, http://www.statistics.gov.uk/cci/article.asp?id=1471 • ONS, Census 2001 Definitions - Output classifications part 2http://www.statistics.gov.uk/statbase/Product.asp?vlnk=12951 • ONS, Guidance on presenting and discussing religion data Addendum to Ethnic Group Statistics: A guide for the collection and classification of ethnicity data, http://www.ons.gov.uk/about-statistics/measuring-equality/ethnic-group-statistics/index.html • Statistics Canada, Ethnic Origin Reference Guide, 2006 Census, http://www12.statcan.ca/census-recensement/2006/ref/rp-guides/ethnic-ethnique-eng.cfm • Statistics Canada, Languages Reference Guide, 2006 Census, http://www12.statcan.ca/census-recensement/2006/ref/rp-guides/lang-eng.cfm • UNECE/CES Recommendations for the 2010 Censuses of Population and Housing, prepared in cooperation with the Statistical Office of the European Communities (Eurostat), 2009, New York and Geneva
Thanks for your attention For any further information please contact: zindato@istat.it