Comparing survey data and administrative records on gross earnings: nonreporting, misreporting, interviewer presence and earnings inequality

Assessing dissimilarity of employment history information from survey and administrative data using sequence analysis techniques

“…Adriaans et al 2020) and income (e.g. Kreiner et al 2015;Valet et al 2019), counteracting the recall bias of isolated life-course events such as retirement (Korbmacher 2014), or decreasing the measurement error associated with entire dimensions of the life course such as union histories (Kreyenfeld and Bastin 2016) and employment biographies (e.g. Huber and Schmucker 2009;Kreuter et al 2010;Wahrendorf et al 2019).…”

“…Korbmacher 2014;Schröder 2011;Wagner and Philip 2019). Alternatively, they may withhold or misrepresent sensitive aspects of their past such as (un-)employment spells, earnings or partnership history due to social desirability bias (Krumpal 2013;Valet et al 2019). The survey design, interview mode and interviewer characteristics may exacerbate or mitigate the likelihood and severity of these biases (Kreuter et al, 2008;Kühne 2018;West and Blom 2017).…”

Gender Differences in Job Resources and Strains in Authority Positions

“…I log-transform the variable to address skew and adjust 2014 and 2015 earnings to 2016 earnings according to the Consumer Price Index. My measure of earnings is based on administrative tax records data, making it much less prone to error than the survey-based earnings estimates used in prior research that underestimate inequality through high levels of underestimates by high earners and overestimates by low earners (Kim and Tamborini 2014; Valet, Adriaans, and Liebig 2019). In addition, earnings in the Dutch registers are not top-coded; this is particularly important for correctly estimating earnings differences, given exceptionally high compensation among some individuals in authority, such as top managers (Kalleberg 2011).…”

Constraints to travel outside the local area: Effect on social participation and self-rated health

“…Income was missing for 195 of the remaining respondents. These respondents were retained in the analysis for two reasons: 1) to prevent data loss on a key variable explaining self-rated health, and 2) to avoid bias, since nonresponse in income questions tends to be related to income levels themselves, which leads to bias when excluding missing data (Valet et al 2019). Household income was first converted into per-person income, based on the number of adults and children in the household.…”