Go to:
Even in the age of big data, social science surveys and panel infrastructures are an essential supplement to official data for social science research and as a basis for empirically sound policy advice. They are representative and cover relevant topics with scientifically validated measurement instruments. In addition, they also explore the attitudes and personality traits of respondents, can document changes in individual life courses as repeat surveys and potentially also uncover causal mechanisms.
However, these science-based infrastructures have faced fundamental challenges in recent years, which also represent opportunities. These include the declining willingness to participate and rising survey costs, but also the possibilities of digitization for surveys, the linking of data sets and the inclusion of multidisciplinary perspectives.
Following an overview of the relevant German survey landscape, the publication on the "Future of social science surveys and panel structures" by Prof. Dr. Gert G. Wagner and Prof. Dr. Martin Bujard discusses the central challenges, development possibilities and opportunities of social science surveys and panels.
Wagner and Bujard note that even the enormous growth in big data cannot replace the tasks of surveys and panel infrastructures. With regard to scientific questions and quality criteria, data must not be distorted, but should be representative of the population as a whole and its (relevant) subgroups. In addition, the data collected must contain the relevant information (variables) for the various scientific questions and must also be valid. Both are almost systematically not the case with data collected for commercial purposes or on the Internet.
Official statistics offer several excellent data sets that are based either on mandatory information (as in the case of the census and microcensus) or on official reports from authorities (e.g. birth statistics). They are based on very high case numbers and also better capture population groups for which traditional surveys have lower response rates (for example, people with a low level of education or a migration background). However, even officially conducted population samples, in which respondents are not subject to any obligation to provide information, have to contend with the same challenges in terms of response rates as all other surveys. The claim that there are generally valid, so to speak objectively correct data surveys can therefore not be met by official surveys either.
Furthermore, it cannot be decided centrally or even by law which questions are scientifically and socially interesting and relevant. Rather, this depends on different interests and assessments.
Social science surveys are always subject to methodological and content-related challenges, as both survey methods and content-related questions are constantly changing. For Wagner and Bujard, it is clear that the development and testing of new content and methods is primarily the task of science. Small and large social science studies will continue to develop independently of each other - and this heterogeneity is a good thing in this case.
In addition, the interviewees must not be overburdened in terms of time. One hour is the upper limit for face-to-face interviews, while web-based surveys should not exceed 20 to 30 minutes per survey session. This limits the content of the survey, from which Wagner and Bujard deduce that there should be various specialized survey programs with different focal points in terms of content (and methodology).
Wagner and Bujard identify non-participation as the greatest methodological challenge in representative voluntary surveys - be it unit non-response (complete non-participation) or item non-response (refusal to answer individual questions).
In recent decades, unit non-response has risen sharply - to around 70% in high-quality scientific surveys (and to over 90% in commercial telephone surveys). The final data set therefore requires a weighting that balances out different participation probabilities as far as possible.
The greatest analysis problems are caused by groups that do not participate in surveys at all or only to a small extent, which in turn leads to small case numbers in unweighted samples that cannot be analyzed. Typically, no attempt is even made to interview homeless people. The same applies to people living in communal accommodation (e.g. retirement homes) or institutions (e.g. prisons). In purely online surveys, the offline population, which is often older people, is systematically excluded. Although migrants are not systematically excluded, the response rates for this population group are often lower than for German-speaking respondents due to language problems.
Wagner and Bujard assume that the effort involved in recruiting respondents for large population-representative samples will tend to increase. This means that there will probably be more special surveys for small and difficult-to-survey groups in the population in the future.
Internet-based survey modes are already being used in several surveys. This trend will increase, also due to the significantly lower costs and comparable quality. In addition, web-based surveys open up highly interesting new possibilities; among other things, the current margin can be reached and thus the timeliness of results improved.
Surveys during the year will be made easier and are likely to increase. Online surveys can be used very well and efficiently both on a regular basis and for current ad hoc surveys. This also applies to event-related online surveys, which can respond promptly to individual events such as unemployment or separation from a partner, for example.
The pool of panel respondents can and will be opened up more and more for qualitative survey methods. This in turn will lead to more texts in the data sets, which can be analyzed using a wide variety of methods.
In addition to online surveys, other forms of data will also play a growing role, the authors predict. These include video material or medically and biologically relevant data (for example, the evaluation of movement profiles using special apps on smartphones). The linking of process-generated data, such as data from electronic patient records, is also likely to increase.
Finally, web-based surveys also enable experimental components in surveys (or in their innovation samples). This makes it comparatively easy to conduct experiments with smaller subgroups of the panel.
In the future, according to Wagner and Bujard, biologically and medically relevant data will increasingly be linked with social science surveys. Behavioral and psychobiological measurements could also be incorporated into studies, for example to investigate questions of well-being with regard to risk and resilience factors.
The authors are rather skeptical about collecting saliva or blood samples to obtain genomic material from respondents, as social science surveys do not have sufficiently large samples for genetic discovery studies.
In contrast, the authors believe that linking microdata to specific individuals will play a greater role in the future. A whole range of information is difficult to collect. This applies in particular to consumption, the use of medical services, but also to the income and mobility behavior of respondents. However, there is currently no legal basis that would allow such links to be made.
However, it is already possible to link survey data with selected process-produced data if the respondents explicitly agree. For example, some surveys are linked to data from the statutory pension insurance scheme and the Federal Employment Agency. However, it can - rightly - be ruled out that there will be a legal situation that would allow all scientifically interesting data sets to be linked to each other without being asked.
Social science surveys and panel infrastructures continue to be highly relevant for basic research and policy advice. "Big data" can in no way replace these, as representative statements can only be achieved through targeted surveys.
Official data from censuses and microcensuses provide a good basis for weighting survey data based on voluntary information from respondents. Internet-based survey forms do not pose a fundamental representativeness problem for the survey. However, as long as certain population groups are very different and also socially selective and cannot be reached via the internet, non-network-based survey modes will continue to be necessary for population-relevant statements.
One of the advantages of digital surveys is that they can be collected and processed promptly for the current margin. However, nowcasts will continue to be necessary for constructs that are difficult to collect and time-consuming to process.
Psychology could become more important for large surveys and panel studies in the future. The addition of some psychological concepts (such as the "Big Five" or control beliefs) have made the prognostic relevance of these psychological variables clear. In addition, the validity of small (student) samples is now highly questioned within psychology. Large surveys are therefore becoming increasingly interesting for the subject.
The selection of relevant new items that enable innovative publications remains the main strategic task for the management of data infrastructures. The competition to identify the most innovative items from several disciplines at an early stage will increase.
The existence of several panel infrastructures and the increasing professionalization and Europeanization will lead to an increase in comparison and benchmarking.
The networking and dissemination of results is of central interest to panel structures. This is why user conferences and training events, as well as press relations and social media, are important tasks for surveys. Close networking with politicians is also necessary in order to find out what is needed and to be able to incorporate current data into political processes.
The complete publication on the "Future of Social Science Surveys and Panel Structures" by Prof. Dr. Gert G. Wagner and Prof. Dr. Martin Bujard can be found here.
Information on a discussion event on the "Future of Social Science Surveys and Panel Infrastructures", which took place as part of the scientific symposium "Daten.Forschung.Zukunft" on July 6, 2023 in Wiesbaden, can be found here. In addition to FReDA project leader Prof. Dr. Martin Bujard, Prof. Dr. Cordula Artelt, Prof. Dr. Gert Wagner and Prof. Dr. Christof Wolf also took part in the event.
Gert G. Wagner stands for the establishment and further development of panel studies in Germany. For decades, he led and shaped the German Socio-Economic Panel Study (SOEP) and developed it further with numerous innovative ideas - making it one of the most important social science panel studies in the world. Gert G. Wagner died on January 28, 2024. The social sciences have lost one of the most important scientists of his generation and the Federal Institute for Population Research one of its first Fellows.
FReDA also benefited from his great expertise. We thank him for this and will continue to be inspired by many of his ideas even after his untimely death.
Cookies help us to provide our services. By using our website you agree that we can use cookies. Read more about our Privacy Policy and visit the following link: Privacy Policy
OK
