Frequently Asked Questions about ALLBUS
1. Does ALLBUS use a recurring set of questions?
All the questions in ALLBUS surveys are intended for replication, but in different ways.
The kinds of replications in ALLBUS are as follows: 1. A set of questions asked in every ALLBUS on background information about respondents and their socio-economic context. These include sex and age, but also household composition, marital status, educational qualifications, and occupation, as well as voting behaviour and interest in politics. 2. Sets of questions on specific topics replicated regularly between every 4-6 years. Examples are questions on awareness of environmental damage, on attitudes towards and contact with foreigners, and on attitudes to marriage and family. 3. Detailed sets of questions on one or two topics per ALLBUS which are replicated about every 10 years.
The kinds of replications in ALLBUS are as follows: 1. A set of questions asked in every ALLBUS on background information about respondents and their socio-economic context. These include sex and age, but also household composition, marital status, educational qualifications, and occupation, as well as voting behaviour and interest in politics. 2. Sets of questions on specific topics replicated regularly between every 4-6 years. Examples are questions on awareness of environmental damage, on attitudes towards and contact with foreigners, and on attitudes to marriage and family. 3. Detailed sets of questions on one or two topics per ALLBUS which are replicated about every 10 years.
2. What type of documentation is available for the ALLBUS surveys?
Documentation materials standardly available include:
- All available original documents used in surveying the data (e.g. questionnaire and additional materials)
- Comprehensive Methodological Reports
- Assiduously processed data sets in different formats (SPSS, Stata and portable) as scientific use files and public use files
- Added value-documentations such as Variable Reports and supplements
3. Which questions have thus far been replicated in ALLBUS surveys?
The ALLBUS codebooks and the cumulative data sets have all the questions which have been replicated at least once.
4. Is ALLBUS a panel study?
No, ALLBUS is not a panel study. It is a time series of cross-sectional surveys.
5. Where can I find research that has used ALLBUS data?
In the ALLBUS bibliography. This contains all the research traced to date using ALLBUS studies.
6. Does ALLBUS data need to be weighted?
It depends on what you want to do with the data. Two kinds of weighting should at least be considered: a design weight to redress for oversampling and a so-called transformation weight to adjust for different probabilities of selection at the level of the individual.
Design weight: The eastern part of Germany, covering the territory of the former GDR, has been oversampled since 1991. If you want to analyse all of Germany together as "the Germans", you need to weight in order to adjust for the oversampling in the eastern states. An East/West weight is provided for this purpose in the data sets. The design is documented (in German) in Gabler (1994) and in the 1994 ALLBUS methods report.
Transformation weight: In surveys where the issued sampling unit is not an individual but a household, you may want to weight to adjust for different probabilities of selection of individuals within households. The 1994, 1996 and 2000 ALLBUS studies have individuals as the issued sampling unit. For other ALLBUS studies, the data sets include a transformation weight. Opinions differ on whether weighting the data necessarily improves measurement, given that the contactibility of households is in part related to household size and composition.
7. What do I need to know about the different types of samples used in ALLBUS studies?
ALLBUS samples are all full probability samples but they differ in design and in the probability of selection of individuals. In studies where the issued sampling unit is an individual, respondents have a known and equal probability of selection. In studies for which the issued sampling unit is a household (1980 through 1992 and 1998) the probability of selection for households is equal, but the probability of selection for individuals varies with the size of household. (It is inversely proportionate to the number of members of the household who are eligible to be interviewed.)
In theory, if analysis is at the level of the individual, ALLBUS data sets with households as the issued sampling unit require to be weighted to adjust for unequal probability of selection at the level of the individual.
In practice, weighting may make little or no difference. The strength of effect is proportionate to the correlation between the "reduced household size" (i.e., those members of the household eligible for the survey) and the characteristics being investigated. We suggest you compare your findings with and without weighting. If there are noticeable differences, we suggest you use the weight. However, when we compared weighted and unweighted findings with Microcensus data, the weighted data were sometimes more skewed.
In theory, if analysis is at the level of the individual, ALLBUS data sets with households as the issued sampling unit require to be weighted to adjust for unequal probability of selection at the level of the individual.
In practice, weighting may make little or no difference. The strength of effect is proportionate to the correlation between the "reduced household size" (i.e., those members of the household eligible for the survey) and the characteristics being investigated. We suggest you compare your findings with and without weighting. If there are noticeable differences, we suggest you use the weight. However, when we compared weighted and unweighted findings with Microcensus data, the weighted data were sometimes more skewed.
8. Why do current ALLBUS data sets no longer use consecutive 'V-numbers' as variable names?
Beginning with ALLBUS 2016, all variables in the ALLBUS-program are assigned a unique and non-arbitrary variable name, which will be persistently used in subsequent cross-sectional and cumulative data sets. For ALLBUS-users this will have the advantage that scripts and syntax files can be used on different data sets without having to adapt variable names to each individual data set. Compared to arbitrary variable names, the use of mnemonic and structured variable names should also make it easier for users to navigate the data.
For a comprehensive description of the rules for naming variables in the ALLBUS-program please refer to the service document Variablennamen in ALLBUS: Die Namenssystematik für Variablen ab ALLBUS 2016 (189 kB).
9. Warum werden ab ALLBUS 2016 Missing Values mit negativen Werten codiert?
Ab ALLBUS 2016 werden fehlende Werte ausschließlich mit Codes aus dem Bereich der negativen ganzen Zahlen codiert. Damit gibt es, anders als bei der bisher verwendeten Konvention, keine Überschneidung der Wertebereiche für valide Werte und fehlende Werte mehr. Der Code '0' z.B. steht jetzt immer für eine valide Antwort und wird nicht mehr für TRIFFT NICHT ZU oder NICHT ERHOBEN verwendet. Außerdem entfällt auch die Notwendigkeit Codes für fehlende Werte an die jeweilige Spaltenbreite der Variable anzupassen (9 / 99 / 999 usw. für KEINE ANGABE). Stattdessen kann jedem Ausfallgrund ein eindeutiger und fester Code zugeordnet werden. Die inhaltliche Zuordnung der neuen Codes orientiert sich soweit möglich an den bisher verwendeten Werten: statt 8 / 98 / 998 für WEISS NICHT wird zukünftig z.B. -8 verwendet und aus 9 / 99 / 999 für KEINE ANGABE wird -9.
Eine ausführliche Darstellung des neuen Missingkonzepts und der verwendeten Codes finden Sie in der Servicepublikation "Kodierung und Definition von fehlenden Werten im ALLBUS: Ein vereinheitlichtes Missing-Schema" (204 kB).
Get in touch
ALLBUS