6 PASS in Transition - Major developments and Innovations.
Contact authors: Beste, Jonas; Collischon, Matthias; Stegmaier, Jens, Wenzig, Claudia
Over nearly two decades, PASS has evolved continuously to address changing social realities, methodological challenges, and innovations in survey research. This chapter highlights the major developments that have shaped the design and implementation of the study, reflecting PASS’s commitment to maintaining methodological excellence, representativeness, and data quality.
Section 6.1 describes the expansion of survey languages and the integration of additional refugee samples to ensure adequate representation of new migrant populations within the SGB II framework. Section 6.2 presents the introduction of a web-based survey mode, designed to explore digital data collection approaches and broaden thematic coverage. Section 6.3 explains the adoption of an adaptive fieldwork design, which employs predictive modelling and targeted incentives to counteract non-response and attrition. Section 6.4 outlines the adjustments implemented during the COVID-19 pandemic, which required rapid adaptations in fieldwork operations, questionnaire design, and interviewer training. Finally, Section 6.5 introduces the interviewer-monitoring system developed for PASS, detailing its integration into CAPI fieldwork and its role in safeguarding data integrity through statistical and procedural quality-control measures.
6.1 Additional Refugee Samples and Adaptations in Survey Languages
The PASS study was established to serve as a central and representative data source for research on basic income security in Germany. To achieve this objective, the composition of the SGB II18 population, including its demographic dynamics, must be continuously monitored and methodologically adjusted where necessary.
Because migrant groups are overrepresented among SGB II recipients, the inclusion of individuals with a migration background has been a methodological priority from the outset. To enable participation of migrants with limited German language skills, the survey was initially administered in Turkish, Russian, and English alongside German language, as these languages reflected the largest linguistic groups among SGB II households at the time. The English version was discontinued after the first wave due to low uptake (nine households). Similarly, the Turkish version was discontinued after wave 9, as participation declined from 163 households in wave 1 to only 14 in wave 9 (see Berg et al. (2010), p. 23).
The large influx of refugees, particularly from Syria since 2015, and from Ukraine since 2022, significantly changed the composition of SGB II recipients. To maintain representativeness and avoid non-response bias, PASS expanded its language offerings. Arabic was introduced in wave 10 to facilitate participation among Syrian and Iraqi refugees. Following Russia’s full-scale invasion of Ukraine in 2022, the share of Ukrainian refugees in the SGB II population rose sharply; as a result, Ukrainian was added in wave 17. Because both Ukrainian and Russian are widely spoken in Ukraine, respondents may choose between these languages during interviews, which has helped sustain high participation rates and methodological consistency. The Table 6.1 summarises the use of foreign language instruments that were introduced in PASS across waves.
This table lists foreign-language questionnaire versions used in PASS, including the wave in which each version was introduced and last used.| Foreign-language version | Introduced in | Last used in |
|---|---|---|
| English | Wave 1 | Wave 1 |
| Turkish | Wave 1 | Wave 9 |
| Russian | Wave 1 | Ongoing (Wave 19) |
| Arabic | Wave 10 | Ongoing (Wave 19) |
| Ukrainian | Wave 17 | Ongoing (Wave 19) |
Foreign-language interviews are primarily conducted via CATI (telephone interviews) by native-speaking interviewers. As most CAPI (face-to-face) interviewers are German-speaking, non-German-speaking respondents are encouraged to transition from CAPI to CATI to improve cost efficiency and ensure data quality (see FDZ-Datenreport 07/2013, User Guide, Ch. 3.3.2, for language use until wave 6). All translations undergo a multi-stage quality assurance process aligned with international standards such as those of the European Social Survey (Dorer, 2018). Since wave 10, survey instruments have been designed to allow interviewers to switch languages at any point during the interview.
To ensure sufficient case numbers for subgroup analyses, Syrian and Iraqi households receiving SGB II were oversampled in the early years following the 2015 refugee arrivals. As their representation in the main sample increased, this oversampling was gradually reduced and later discontinued. For Ukrainian refugees, oversampling was unnecessary, as they were adequately represented through the new entries sample. All Syrian and Ukrainian refugee households are identifiable in the dataset via dedicated sample variables.
The inclusion of refugees as a relevant sub population required additional methodological adjustments beyond language expansion. These included the development of targeted questionnaire modules (e.g. recognition of foreign qualifications), measures to ensure comparability across language versions (German and foreign-language interviews) and interview modes (CAPI and CATI). Cultural sensitivities in interview conduct and new approaches to panel maintenance also played a crucial role. As part of panel maintenance, efforts were intensified to collect stable contact information (email addresses, social media accounts, third-party phone numbers) to counteract the high mobility of refugees, who frequently changed residence after arrival due to administrative relocation or personal decisions.
In summary, the integration of refugees represents one of the most significant methodological adaptations in the 20-year history of PASS. These adjustments in sampling, language provision, and survey methods have significantly enhanced the representativeness of the survey, providing a robust basis for analysing both SGB II dependency and the social and economic integration of refugees in Germany.
6.2 Web survey
To gain experience with online data collection and to provide an opportunity for collecting information on specific topics that were too extensive for inclusion in the main PASS survey, a web-based survey was introduced following wave 16. All respondents from the CAPI and CATI main survey were invited to participate.
The web survey is conducted in autumn each year, no later than six months after the main interviews (but in some cases even sooner). To date, it has been implemented following PASS waves 16 and 18. Unlike the main interviews, which are administered via telephone or face-to-face, the web survey is self-administered. Each wave is designed to last approximately 10–20 minutes, and participants receive a post-paid incentive of 5 euros, sent by post upon completion of the questionnaire.
The web survey questionnaire contains no standard recurring modules; rather, it focuses on special topics. The survey following wave 16 included questions on sanctions for individuals receiving unemployment benefits, employment, leisure activities, and political participation. It also repeated selected questions from the main survey to enable analyses of survey-mode effects (Mühlbacher & Trappmann, 2024). The survey following wave 18 included an extended household deprivation module compared to the main questionnaire, as well as a survey experiment on unemployment-benefit uptake.
Further details on the design, content, and implementation of the web survey are provided in the corresponding data reports (Dickmann et al., 2023).
6.3 Adaptive Design in PASS
Ensuring high data quality and minimising attrition are central objectives of the PASS. Beginning in wave 7 and standardised from wave 14 onwards, PASS introduced an adaptive design approach to fieldwork aimed at improving response rates and reducing bias by prioritising cases with a high risk of non-response. This innovation marked a shift towards a more targeted, data-driven fieldwork strategy that has since been integrated into standard survey operations (for further methodological details and evaluation results, (see Beste et al. (2023)).
The foundation for adaptive design in PASS lies in the extensive longitudinal data accumulated over previous waves. These data include survey responses, para-data, and interviewer assessments, which enable the PASS team to predict the response propensity of each household. While protocols of the adaptive design changed often between waves 7 and 13 (cf. Trappmann & Müller (2014), Kreuter et al. (2014)), the standardized approach that is currently used, was first implemented in wave 14 as part of an experimental study, in which households with a low predicted likelihood of participation were randomly assigned to receive increased prepaid incentives (20 euros instead of the standard 10 euros). The aim was to assess whether targeted incentives could reduce drop-out rates and mitigate non-response bias.
Prediction models were developed using machine learning algorithms trained on data from waves 4 to 12 and validated against wave 13. Random forest models proved particularly effective in identifying households at high risk of attrition. In wave 14, these predictions informed case prioritisation and treatment assignments in both CATI and CAPI modes. The experimental results demonstrated that doubling the incentive significantly increased participation among low-propensity households and yielded modest improvements in overall data representativeness (Beste et al., 2023).
Following these successful results, adaptive design became a regular component of PASS fieldwork from wave 16 onwards. In place of randomised experimentation, the prioritisation process is now systematically applied using updated prediction models based on the most recent data. Households flagged as low propensity are routinely targeted with enhanced fieldwork measures, most notably through increased incentives.
This transition to a standard operational practice illustrates PASS’s commitment to evidence-based survey methodology and continuous quality improvement. The adaptive design helps preserve the longitudinal integrity of the panel by counteracting attrition in key subgroups. It also ensures a more efficient allocation of fieldwork resources by targeting efforts where they are most needed.
For users of the PASS SUF, the integration of adaptive design strengthens both the reliability and validity of the data. The prioritisation process is carefully documented, and its effects on fieldwork outcomes are regularly evaluated. Although the current focus is on targeted incentive adjustments, the adaptive design framework allows for future expansion to other targeted measures, such as interviewer assignments or tailored contact strategies.
6.4 Impacts of COVID-19 Pandemic on Fieldwork, Questionnaire, Interviewer Training, and Other Areas
The COVID-19 pandemic (2020–2022) had a profound impact on the PASS survey, particularly during wave 14, when numerous methodological and content-related adjustments were necessary to respond to the exceptional circumstances while safeguarding data quality and maintaining panel stability. Wave 15 was still heavily influenced by the pandemic, while wave 16 marked the gradual return to standard survey operations. By wave 17, PASS had fully resumed its regular routines.
This section outlines the major pandemic-related adaptations in fieldwork, questionnaire content, and interviewer training. A detailed discussion of the same follow after the Table 6.2.
This table compares the impact of the COVID-19 pandemic on PASS across Waves 14 to 16, covering pandemic events, fieldwork and survey mode, questionnaire changes, interviewer training, and other study areas.| Category | Wave 14 (2020) | Wave 15 (2021) | Wave 16 (2022) |
|---|---|---|---|
| Pandemic events | First nationwide lockdown (16 March 2020); school closures until May; gradual easing from May; stable summer period | Second lockdown (winter 2020/21); gradual reopening from Spring 2021; schools operating with restrictions | Restrictions lifted from spring 2022; partial mask mandates; high vaccination rate |
| Fieldwork / Survey mode | Stoppage of CAPI fieldwork and transfer to CATI; all CAPI halted mid-March; introduction of “CAPI by phone” | “CAPI by phone” fully replaces face-to-face interviews; used as a survey mode from start | Partial return to on-site CAPI (Jun 2021); full resumption by May 2022; “CAPI by phone” optional |
| Questionnaire | COVID-related updates introduced 8 April 2020: home isolation & quarantine (HCV0100, HCV0200); childcare (HKI0900–HKI01100); employment (PCV0200–PCV0600); living situation (PCV0100); mental health (PG1215*) | Continuation of COVID-specific questions | Continuation/adaptation of COVID-specific modules; later removal |
| Interviewer training | Traditional in-person training; after lockdown: webinars; training for “CAPI per phone” | Full shift to online training; digital training concept developed | Consolidation of digital training; continued use of webinars and online modules |
| Other areas | Fieldwork extended until 29 Sept 2020; suspension of interviewer caps | Expanded refreshment sample; oversampling of self-employed SGB II recipients | CAPI vignette module implemented as separate online survey |
6.4.1 Fieldwork changes
Preparations for wave 14, including questionnaire development, sampling, household contact strategies, and interviewer training were carried out under standard PASS procedures. Fieldwork began on 14 February 2020, but the nationwide lock down on 16 March 2020 brought abrupt and far-reaching changes affecting both private and professional life as well as operational fieldwork.
To avoid drop-outs caused by the sudden replacement of interviewers, especially among long-standing panel households previously interviewed in-person, the “CAPI by phone” method was introduced in mid-May 2020. This allowed former CAPI interviewers to conduct interviews by telephone, maintaining interviewer’s continuity and preserving rapport with respondents, albeit in an auditory format. We are aware that the name may be misleading as CAPI by phone interviews are telephone interviews and not personal interview. However, they differ from regular CATI interviews in that they are not administered from a centralized call-centre with regular supervision and are typically administered by CAPI interviewers who previously interviewed the same households face-to-face. In this sense they are hybrids between CATI and CAPI.
To address reduced accessibility, the field period for wave 14 was extended until 29 September 2020. In addition, the usual interview caps per interviewer (CATI: 100 interviews; CAPI: 40 households) were suspended for wave 14 and handled more flexibly in subsequent waves (see Method reports wave 14 (Beste et al., 2021, p. 129); 15 (Jesske et al., 2022, p. 137); 16 (Jesske & Gerber, 2023, p. 128)).
Fieldwork in waves 15 and 16 remained affected by pandemic-related restrictions (see Method report wave 15 (Jesske et al., 2022, Ch. 5.1); 16 (Jesske & Gerber, 2023, Ch 5.1)). In wave 15, in-person contact with households was prohibited at the start of the field period, and “CAPI by phone” served as the full substitute for face-to-face interviews. A vignette experiment on maternal employment and childcare, originally planned as a CAPI module, was implemented as a separate online survey (Frodermann et al., 2024); and (Jesske et al., 2022, Ch. 5.8).
From June/July 2021, in-person household interviews were gradually
reintroduced under strict hygiene protocols, though “CAPI by phone”
remained vital for established panel households. It was only in wave 16
that in-person interviewing resumed widely. From May 2022, CAPI
interviews were once more conducted on-site, with “CAPI by phone”
retained as a complementary method for hard-to-reach households. In wave
17, PASS fully returned to standard procedures, with CAPI reserved for
in-person interviews and CATI for telephone interviews (with very few
exceptions).
To compensate for drop-outs and reduced case numbers in wave 14, the
sampling strategy was adjusted. In wave 15, the refreshment sample was
expanded (see Berg et al. (2023), p. 18). Moreover, the sample was extended to
targeted subgroups, notably self-employed persons receiving SGB II
benefits, whose numbers had increased sharply during the pandemic.
Oversampling enabled detailed subgroup analyses.
6.4.2 Questionnaire Adjustments
In addition to methodological adaptations, wave 14 introduced new modules to capture the realities of life during the pandemic (see Berg et al. (2022), Ch. 1.2). From 8 April 2020 onwards, pandemic-specific questions supplemented both household and person questionnaires.
Household-level topics included the extent of household affectedness (e.g., home isolation or quarantine) and the organisation of childcare during the pandemic. The personal questionnaire included items on employment (short-time work, job loss, remote work), life changes, and mental health.
Certain items were accompanied by interviewer instructions clarifying whether responses should refer to pre-pandemic or current conditions. These changes during ongoing fieldwork were implemented across all language versions (German, Arabic, and Russian).
The variable kennungfbversW14 allows users to distinguish interviews conducted before and after the introduction of pandemic-related questions and interviewer instructions. In waves 15 and 16, most COVID-19 questions were continued or modified to reflect evolving conditions, though the home isolation/quarantine module was discontinued after wave 15. Since wave 17, all pandemic-related questions have been removed from the questionnaire.
6.4.3 Interviewer Training
Interviewer training also underwent major transformations. While wave 14 training initially took place in person, all subsequent sessions after the lockdown were moved online. This ensured that interviewing quality remained high under restrictive conditions.
In waves 15 and 16, a fully digital training concept was further developed (Jesske et al., 2022, Ch. 7.1). Training materials were provided as self-learning modules with integrated knowledge checks, and interactive webinars reinforced the content. Digital training has since become a permanent feature of PASS interviewer preparation.
6.4.4 Potential Implications for Data Use and Analysis
Through early and coordinated measures by the IAB and the fieldwork institute, PASS successfully adapted to the challenges posed by the COVID-19 pandemic between 2020 and 2022, maintaining high-quality data and panel stability.
Key to this success were the flexible survey design enabling a rapid shift to telephone interviewing (CATI or CAPI per phone) and the timely inclusion of pandemic-relevant questionnaire content. Flexible sampling strategies targeting affected subgroups and digital interviewer training ensured robust and unbiased data collection.
For data users analysing waves 14 to 16, it is essential to consider these methodological and content-related adaptations systematically. Relevant features, particularly “CAPI per phone” as a supplementary survey mode, and the wave 14 questionnaire version should be accounted for in analyses. Pandemic-related changes affected multiple aspects of data collection, including interview duration, contact quality, interviewer continuity, and response behaviour. These factors may influence data comparability across waves and should ideally be controlled for, in longitudinal analyses.
6.5 Interviewer Monitoring in the PASS Study: Integration into CAPI Fieldwork
Maintaining high data quality is a core principle of PASS. In recent years, interviewer monitoring has become an increasingly important component of this effort. As interviewer-administered surveys can be vulnerable to both intentional falsification and deviations from the prescribed standardised interview procedure, PASS has developed a systematic approach to identify and address such risks - particularly within its CAPI fieldwork.
In contrast to CATI (Computer-Assisted Telephone Interviewing), where interviewers operate from supervised call centres with real-time oversight, CAPI interviewers work independently in the field. This autonomy limits direct insight into interview conduct and creates the need for dedicated quality-assurance mechanisms tailored to CAPI’s decentralised nature.
Audio recordings of all CAPI interviews have been collected (given respondent consent), since wave 10, providing valuable material for monitoring and for understanding interview dynamics in the field. An interviewer’s recording rate additionally serves as an indicator of adherence to survey protocols.
A comprehensive monitoring strategy was piloted following the field period of wave 14 and throughout wave 15. From wave 16 onwards, interviewer checks have been embedded directly into the data collection process. These checks allow for the timely identification of suspicious behaviour and enable targeted actions, such as clarifying anomalies, retraining interviewers, excluding them from fieldwork, or initiating follow-up interviews when necessary. This proactive approach prevents further data collection by problematic interviewers and minimises the risk of error propagation across waves.
The integration of interviewer monitoring into PASS represents a significant step towards safeguarding data quality in decentralised CAPI settings. It enables structured, evidence-based responses to potential irregularities and complements existing quality controls in CATI. As a result, users of the PASS SUF can rely on data that benefit from both robust preventive measures and an active commitment to continuous quality assurance.
The quality-control strategy in PASS combines statistical diagnostics and procedural checks to ensure comprehensive oversight (Beste et al., 2021):
Analysis of timestamps:
Timestamps recorded during interviews are analysed to detect unusually short durations at the item or module level, which may indicate skipped questions or interviewer-entered responses.
Duration comparisons:
Interviews with and without audio recordings are compared, as interviewers may alter their behaviour when aware that a session is being recorded. Suspicious deviations in timing can therefore reveal irregularities.
Longitudinal consistency checks:
For relatively stable variables (e.g., political attitudes), large deviations across waves can signal data irregularities or potential falsification.
Statistical indicators:
PASS employs a set of 23 statistical indicators designed to identify anomalies in response patterns (e.g., high rates of item non-response, violations of Benford’s law, non-differentiation). A meta-indicator (Schwanhäuser, 2024) aggregates these measures at the interviewer level to flag unusual data characteristics.
Procedural checks by infas (the survey institute):
These include systematic evaluations of audio recordings, postal validations with respondents, and targeted re-contacts to verify interview authenticity.
Overall, the monitoring system focuses not only on detecting outright falsifications, but also on identifying partial falsification and non-standardised interview conduct, both of which can compromise data integrity.
Interviewer monitoring is carried out in close collaboration between the PASS study team at the IAB and the survey institute infas, ensuring continuous exchange and alignment throughout the fieldwork process.
The implementation of interviewer monitoring has already led to measurable improvements in CAPI quality assurance. As detailed by Beste et al. (2021), the first applications of the system in waves 14 and 15 resulted in the exclusion of two interviewers and their associated interviews from the dataset due to clear protocol violations.