Creating comparable sociodemographics across German surveys without losing your mind – powered by KonsortSWD

Every survey of individual respondents collects sociodemographic information. Variables such as age, gender, education, income, immigrant background, or employment status are among the most frequently used in empirical research. Yet, these variables are often collected using different survey questions, response options, or coding schemes. As a result, data from different survey studies cannot easily be combined or compared. Sociodemographic standard variables, developed within the KonsortSWD project StanDem, address this challenge via output harmonisation. Rather than recommending how surveys should ask sociodemographic questions, standard variables provide a common target format for converting existing variables from different surveys.

In this blog post, we explain why harmonisation is becoming increasingly important, why sociodemographic standard variables are needed in addition to standardised questionnaire items, and how StanDem supports research data centres (RDCs), survey programmes and researchers in implementing these variables.

Bei jeder Befragung einzelner Teilnehmenden werden soziodemografische Daten erhoben. Variablen wie Alter, Geschlecht, Bildungsstand, Einkommen, Migrationshintergrund oder Beschäftigungsstatus gehören zu den in der empirischen Forschung am häufigsten verwendeten. Allerdings werden diese Variablen oft mithilfe unterschiedlicher Fragen, Antwortmöglichkeiten oder Kodierungsschemata erfasst. Infolgedessen lassen sich Daten aus verschiedenen Erhebungen nicht ohne Weiteres zusammenführen oder vergleichen. Soziodemografische Standardvariablen, die im Rahmen des KonsortSWD-Projekts StanDem entwickelt wurden, begegnen dieser Herausforderung durch eine Harmonisierung der Ausgabedaten. Anstatt Empfehlungen dafür zu geben, wie soziodemografische Fragen in Erhebungen gestellt werden sollten, bieten Standardvariablen ein gemeinsames Zielformat für die Umwandlung bestehender Variablen aus verschiedenen Erhebungen.

In diesem Blogbeitrag erläutern wir, warum Harmonisierung zunehmend an Bedeutung gewinnt, warum soziodemografische Standardvariablen zusätzlich zu standardisierten Fragebogenitems benötigt werden und wie StanDem Forschungsdatenzentren (RDCs), Erhebungsprogramme und Forscher bei der Umsetzung dieser Variablen unterstützt.

DOI: 10.34879/gesisblog.2026.129

---

Why harmonise survey data?

Imagine two surveys that both measure respondents’ marital status. Comparing the data appears straightforward – they both collect information on the same concept. Easy! Looking more closely, however, reveals a different picture: One survey used four response categories for respondents’ marital status, the other used seven. How can a researcher analyse the two datasets together, e.g. to study divorced people – a small group that can be better studied by combining different samples? This challenge extends far beyond marital status, where it may be quite obvious which response categories correspond to each other. Nearly every sociodemographic characteristic – including education, employment status, household composition, immigrant background, citizenship, or household income and gender – is measured differently across studies. For data users interested in merging and combining different surveys, making sociodemographic information comparable not only requires a great deal of effort that goes far beyond writing the required code, but it often also leads to lower measurement quality than the original variables. Lack of comparability also poses a major obstacle to research more generally. In the worst-case scenario, innovative ideas and research questions may not even be pursued further!

The Tandem of input and output harmonisation

How to make survey data comparable? The first solution is input harmonisation. Here, different surveys use identical measurement instruments from the outset. Different surveys ask the same questions, use the same response options or scale and thus collect information in a standardised way. The primary goal is to ensure that different studies measure identical concepts in the same way. Input harmonisation is an effective way of ensuring measurement comparability. In practice, however, it is often difficult to do. Especially, long-running panel and trend studies cannot easily change questions used in earlier rounds of data collection without compromising comparability over time. This is where output harmonisation becomes essential. Instead of using standardised questionnaire items, output harmonisation renders variables comparable after the data have been collected. Conceptually comparable variables are translated into a common coding scheme or scale, making them directly comparable across studies while allowing individual surveys to keep their own questionnaire items.

Standard items are different from standard variables

In Germany, the “Demographic Standards” provide widely recognised recommendations to input-harmonise sociodemographics using standard questionnaire items. The concepts of standard items and standard variables may, at first glance, appear interchangeable: ultimately, we use questionnaire items to obtain variables in a dataset. Do not fall into this trap! Standard survey items are implemented (or more often, not implemented) during the survey design phase. Standard variables, in contrast, are (like other derived variables) implemented during the survey data curation phase to augment the analytic potential and reusability of the data. Rather than prescribing how surveys should ask questions and thus interfering in the survey design phase (which is often characterised by high time pressure), they provide target variables and define how conceptually equivalent information should be represented within a dataset.

Latent constructs vs. sociodemographics

Our colleague Ranjit Singh wrote a whole series of blog posts on the benefits, obstacles and pitfalls(!) of ex-post harmonisation generally, and on latent variable harmonisation specifically. The psychometric methods described by Singh are, however, inapplicable to sociodemographic characteristics: The latter largely represent more or less manifest constructs measured with just one or a small series of complementary categorical questionnaire items (e.g. on employment status). To support the ex-post harmonisation of sociodemographics across German surveys, the KonsortSWD project StanDem developed a set of sociodemographic standard variables, based on an extensive review of measurement instruments used in large German survey programmes and a comprehensive validation study, covering a wide range of concepts such as sex, age, country of birth, education, main status and marital status. They were designed to be derived from a wide range of study-specific questionnaire items and the existing “Demographic Standards”, which, by the way, neither suggest variable names or labels nor values or value labels for the recommended questionnaire items that could be used in the resulting dataset.

What do the sociodemographic standard variables look like?

A central goal of the sociodemographic standard variables is to ensure that conceptually equivalent information appears consistently across datasets. To achieve this, all sociodemographic standard variables follow common specifications. Of course, there must be space for plurality, complexity, and flexibility. For this reason, some sociodemographic standard variables have been designed with minimal and maximal versions. Rather than creating separate variables, different levels of detail are represented in a hierarchical coding scheme. This approach maximises comparability while preserving valuable information wherever possible. For some concepts, alternative coding standards also exist (for example: International Standard Classification of Occupations (ISCO) and the German Classification of Occupation (KldB)).

Turning standard variables into research infrastructure

Developing sociodemographic standard variables is only the first step towards more comparable survey data. Rather than leaving the effort of harmonisation to individual researchers, StanDem aims to shift this work upstream to the organisations that curate and disseminate survey data. Their full potential can only be realised once they are systematically implemented in RDCs and survey programmes and made available to data users. StanDem therefore supports research data centres and survey programmes in implementing the sociodemographic standard variables in their own datasets and publishing them for the scientific community. This approach offers several advantages. First, researchers no longer need to repeatedly derive variables themselves, thereby increasing research efficiency. Second, harmonisation decisions become transparent, consistent, and reusable across research projects. Finally, as more studies adopt the standard variables, the benefits of combining data from different surveys grow and grow.

A structured workflow for implementing sociodemographic standard variables

Implementing harmonised variables across survey datasets requires more than simply writing code. Measurement instruments need to be scrutinised and understood, correspondences established, potential ambiguities resolved, decisions documented, and implementation steps coordinated. To support this process, StanDem has developed a structured implementation workflow. The workflow builds on two pilot implementations carried out in the RDCs of ALLBUS (General German Social Survey) and NEPS (National Educational Panel Study), where the implementation process was tested, refined and translated into a practical guide for future adoptions. It consists of four consecutive phases: kick-off, implementation and documentation, debriefing and feedback, and launch. To make implementation as straightforward and transparent as possible, StanDem offers a range of supporting materials that go with the workflow. The overarching goal is for data producers to eventually implement the sociodemographic standard variables independently, following a transparent and well-documented process and only calling for support in difficult situations.

What has been achieved so far?

Okay guys, that’s all well and good. You’ve developed a nice standard. But what’s the point if no one follows your standard? That’s true. Or let’s say, that was true. The implementation workflow is no longer merely a conceptual framework! Its development has been informed by pilot implementations in the ALLBUS 2021 and the NEPS SC6. These pilot projects demonstrated that the implementation process can be standardised while remaining sufficiently flexible to accommodate study-specific characteristics. The resulting workflow, templates and documentation materials now provide RDCs and survey programmes with practical guidance for implementing sociodemographic standard variables in a structured and transparent way. Most importantly, they lay the foundation for harmonised sociodemographic information becoming a regular component of future survey data releases and no longer remain an additional and sweaty task left to individual researchers. In the coming months, the Socio-economic Panel (SOEP) will implement StanDem’s sociodemographic standard variables. Publication is expected in late spring 2027. We call other RDCs to plan the implementation of sociodemographic standard variables in the data they curate, especially if they cover the general adult population in Germany and are thus suitable for integration with ALLBUS, SOEP, or NEPS SC6. Interested? Get in touch!