A gender gap in gender gaps: social norms and housework reporting

Acht, Martin; Rebaudo, Mara Article — Published Version A gender gap in gender gaps: social norms and housework reporting Empirical Economics Provided in Cooperation with: Springer Nature Suggested Citation: Acht, Martin; Rebaudo, Mara (2025) : A gender gap in gender gaps: social norms and housework reporting, Empirical Economics, ISSN 1435-8921, Springer, Berlin, Heidelberg, Vol. 68, Iss. 6, pp. 2977-3029, https://doi.org/10.1007/s00181-024-02710-z This Version is available at: https://hdl.handle.net/10419/323197 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Empirical Economics (2025) 68:2977–3029 https://doi.org/10.1007/s00181-024-02710-z A gender gap in gender gaps: social norms and housework reporting Martin Acht1,3 ·Mara Rebaudo1,2 Received: 16 May 2024 / Accepted: 27 December 2024 / Published online: 6 February 2025 © The Author(s) 2025 Abstract Gender differences in the amount of housework performed and the role of social norms in explaining these persistent gaps have received increasing attention from both policymakers and researchers in recent years. However, norms may not only affect the actual division of housework but also potentially influence the reporting behavior in surveys. We study how retrospective responses about time-use in faceto-face interviews are influenced by the gender of the interviewer. Our findings show that women tend to report significantly more hours of housework when interviewed by a woman rather than by a man. This effect is not observable for male respondents, resulting in an interviewer gender gap in the housework gender gap. Exploring the effect in relation to several norm-related characteristics indicates that social norms play an important role in the reporting of housework hours. Therefore, gender gap estimates based on face-to-face interviews should be interpreted with great caution. Keywords Gender ·Gender gap ·Doing gender ·Social norms ·Interviewer effects JEL classification D13 ·J16 ·D90 1 Introduction The role of social norms as important drivers of human behavior has attracted the attention of researchers not least since the seminal work by Akerlof and Kranton (2000). In this field of research, a growing literature seeks to understand how social The authors want to thank seminar participants at the EEA Congress, at the Annual Conference of the Verein für Socialpolitik, at the RGS Doctoral Conference, Marianna Schaubert and seminar participants in Sankt Augustin, Freiburg and Göttingen for helpful comments and suggestions. BMara Rebaudo [email protected].de 1Fraunhofer Institute for Applied Information Technology FIT, Sankt Augustin, Germany 2University of Freiburg, Freiburg im Breisgau, Germany 3Federal Ministry of Finance, Berlin, Germany 123 2978 M.Acht,M.Rebaudo norms contribute to explaining persistent gender gaps in earnings, time-use and other outcomes. In their prominent contribution, Bertrand et al. (2015) demonstrate how couples engage in norm-conforming behavior by adjusting their labor market outcomes and time spent on housework to adhere to the traditional male breadwinner norm. In this study, we offer a new perspective on these findings and analyze whether and how social norms, which we define here as socially prescribed gender norms, affect the reporting of housework hours in surveys. Our approach exploits a contextual factor potentially related to the role of social norms in the response process: the gender of the interviewer. Specifically, we estimate respondent-fixed-effects models to study how the reporting of housework hours changes when the interviewer is female as opposed to male. When studying time spent on housework, most evaluations rely on retrospective questionnaire responses rather than on diary-based data. There remains, however, discussion about how reliable such retrospective time-use data are. For example, Kan (2008) finds systematic differences between questionnaire and diary data on housework hours which vary by gender and other characteristics of the respondents. Several mechanisms to explain the gap between questionnaire-based and diary-based estimates regarding housework hours have been proposed in the literature. On the one hand, questionnaire-based estimates are likely to be biased due to an unclear definition of housework, recalling problems or double-counting when housework is performed simultaneously with other activities (Kan 2008; Press and Townsley 1998; Warner 1986). On the other hand, besides these more technical explanations, reporting housework hours in surveys is sometimes considered a gendered process. Press and Townsley (1998) argue that social desirability plays a central role in explaining differences between questionnaire and diary estimates, because some respondents may feel the pressure to report a level of housework that complies with normative gender roles. Our paper focuses on this aspect of housework reporting and is, to the best of our knowledge, the first to provide empirical evidence for whether and how norms affect the reporting of housework hours. A growing body of literature points to the importance of social norms in explaining persistent gender differences in time spent on housework that cannot be rationalized within standard economic models. This relates to findings about how couples react to labor market events (Foster and Stratton 2018), to child birth or parental leave (Schober and Zoch 2019; Schober 2013), to entering retirement (Leopold and Skopek 2015), and also to the role of institutions in affecting gender differences (Cooke 2007; Lippmann et al. 2020). One of the most striking of these observations has been demonstrated by Bertrand et al. (2015): Standard economic models of decision-making within the family would predict that as a spouse’s contribution to the couple’s income increases, she should decrease her housework hours. However, Bertrand et al. (2015) show that women who outearn their partners and thereby violate the male breadwinner norm, actually increase their time spent on housework. This phenomenon has been interpreted in light of the “doing gender” theory first proposed by West and Zimmerman (1987). This theory posits that individuals “do gender” by actively conforming to societal expectations and norms associated with their perceived gender identity. When a traditional gender role is violated, for instance, when a wife outearns her husband, 123 A gender gap in gender gaps… 2979 individuals aim to compensate for this violation by reinforcing their gender identity in other ways, such as increasing their housework hours. The presence of “doing gender” behavior has been established for several countries and contexts (e.g., Bertrand et al. 2015; Lippmann et al. 2020; Bittman et al. 2003; Foster and Stratton 2018). The finding that couples engage in norm-conforming behavior is not only relevant to the gendered division of housework hours, but also extends to women’s labor force participation and income (Bertrand et al. 2015). However, recent studies by Roth and Slotwinski (2020) and Murray-Close and Heggeness (2019) suggest that the observed norm-conforming behavior in this domain is the result of systematic misreporting of incomes. In light of this, while “doing gender” behavior regarding housework hours may indeed be present, it could be in terms of overreporting of housework hours rather than actual behavior (Sullivan 2011). Then, the aforementioned strategies to adhere to social norms may still be relevant, but probably in a different way than the literature currently suggests. Analogous to the work by Roth and Slotwinski (2020), who compare official and reported income data for the same individuals, it would be ideal to compare retrospective responses about time-use to the “true” values to see if certain characteristics lead individuals to overor underreport housework hours. As such data are not available and the credibility even of diary-based estimates remains questionable, we apply a different approach to study how social norms affect the reporting in interviews. Using data from the German Socio-Economic Panel, we study how retrospective responses about time-use in face-to-face interviews are influenced by the gender of the interviewer. If social norms play a role in the reporting of time-use, it seems likely that the characteristics of the interviewer, especially the interviewer’s gender, affect in which direction social norms influence the time-use estimates and how strongly. We focus on routine household chores such as washing, cooking and cleaning, as they are considered among the least enjoyable and also the most time consuming non-market activities (Sullivan 2013; Coltrane 2000), which points to the importance of social norms in explaining persistent gender differences in time spent on these tasks. For other activities the connection to social norms is less straightforward. We find that female respondents report significantly more hours of housework when interviewed by a woman instead of a man. A similar gender-of-interviewer effect for male respondents is not observable. This in turn leads to an interviewer gender gap in the housework gender gap, because women and men respond differently to the interviewer’s gender. We interpret the findings as evidence for the fact that social desirability concerns – and hence social norms – play an important role in the reporting of housework hours. By comparing couples to singles, and also single-to-couple transitions over time, we disentangle the mechanisms behind socially desirable reporting, namely subconscious self-deception versus impression management. Stratifying the sample and interacting the interviewer’s gender with several norm-related characteristics, including respondents’ and interviewers’ cohorts, support the hypothesis that the observed effects are related to traditional social norms. We analyze differences by women’s relative income and employment status and thereby discuss how the interviewer’s gender is related to the finding by Bertrand et al. (2015) that couples “do gender” when women outearn their partners. Following the literature on differences between West and East Germans regarding social norms (Alesina and Fuchs-Schündeln 2007; Bauernschuster 123 2980 M.Acht,M.Rebaudo and Rainer 2012; Beblo and Görges 2018), we find that gender-of-interviewer effects are less present for East Germans. Furthermore, we study differences in the genderof-interviewer effect by source country gender equality of immigrants based on the literature on social norms and cultural heritage (Alesina et al. 2013; Blau et al. 2020; Giuliano 2020). In line with our hypothesis, the effects are smaller for immigrants from more gender-equal countries. We provide the first causal evidence of how housework reports differ by the interviewer’s gender, and how this gender-of-interviewer effect is related to social norms. Therefore, our paper contributes to both the strand of literature on interviewer effects and that on norms affecting the division of housework hours. We find differences in reporting behavior between female and male respondents, indicating that gender gap estimates based on face-to-face interviews should be interpreted with caution. We argue that the extent of gendered reporting is directly related to the internalization of traditional social norms. Thereby, we offer a new perspective on previous findings about the relationship between social norms and time spent on housework, as these may be influenced by reporting effects. While our approach focuses on reporting differences toward female versus male interviewers, our findings suggest that social norms may influence response behavior in even more ways than could be analyzed in this study. The paper is structured as follows. Section 2summarizes the existing evidence on gender-of-interviewer effects and provides a theoretical overview of the concept of social desirability bias to explain the findings. Section3describes the data and Sect.4presents the empirical strategy. Section 5studies gender-of-interviewer effects on housework hours and relates the findings to several norm-related characteristics, including cohorts, employment status and income, differences between West and East Germany and the cultural heritage of immigrants. Section6provides robustness checks to address if any other factors other than the interviewer’s gender are driving the results, before Sect.7concludes. 2 Background Before summarizing the existing literature on gender-of-interviewer effects, we first provide a theoretical overview of the concept of social desirability bias, which provides the foundation for the interpretation of our results. The concept of socially desirable reporting suggests that respondents reflect on societal views and expectations regarding certain topics and adapt their responses accordingly (see Krumpal 2013, for an overview). Hence, instead of answering accurately and truthfully, some respondents distort their answers to conform to social norms and thereby maintain a socially favorable self-presentation. In fact, empirical studies have shown that respondents overreport socially desirable behavior, such as charity donations, voting, or seat belt use, and underreport socially undesirable behavior, such as smoking or illicit drug use (Tourangeau et al. 2000; Tourangeau and Yan 2007; Krumpal 2013). The underlying processes and exact mechanisms of socially desirable reporting are still insufficiently understood (Tourangeau and Yan 2007; Holtgraves 2004). Scholars often distinguish between two types of norm-conforming reporting: Other-deception 123 A gender gap in gender gaps… 2981 or impression management, on the one hand, relates to the individuals’ need for social approval by others, e.g., the interviewer (Paulhus 1991). It assumes that some respondents purposefully tailor their answers to avoid social disapproval or to create a positive image of themselves. Self-deception, on the other hand, refers to individuals’ strategies to reduce cognitive dissonance caused by a divergence between social norms and reality (Krumpal 2013). As opposed to impression management, self-deception is mostly considered to be subconscious and automatic (Tourangeau and Yan 2007). While impression management behavior is directed at other subjects, e.g., interviewers, the main addressees of self-deceptive reporting behavior are the respondents themselves. Importantly, independent of the channel, socially desirable reporting will only occur when a deviation between the honest response and what is perceived to be desirable exists. It has been shown that social desirability bias is especially likely in personal interviews and that interviewer characteristics, such as the interviewer’s gender, can affect the likelihood of engaging in norm-conforming reporting behavior (Tourangeau et al. 2000; Tourangeau and Yan 2007). In terms of impression management, respondents may perceive female and male interviewers to have different opinions regarding gender-related questions and adjust their responses to gain social approval of the interviewer. However, even if respondents were not trying to impress their counterparts, the interviewer’s gender can still affect the response in terms of self-deceptive behavior. Specifically, respondents may be more or less aware of their own deviation from social norms depending on their counterpart’s gender. Following this concept of socially desirable reporting, several studies analyze the role of gender-of-interviewer effects (see West and Blom 2017, for an overview). Most of the previous research has focused on normative views, for example about marriage (Liu and Stainback 2013) or women’s issues and gender equality (Huddy et al. 1997). A common finding in the literature is that respondents either provide more feminist or liberal responses toward female interviewers or, the other way around, more traditional ones toward male interviewers (Kane and Macaulay 1993; Lueptow et al. 1990; Flores-Macias and Lawson 2008; Lipps and Lutz 2017). It is important to note that previous studies on gender-of-interviewer effects almost exclusively rely on cross-sectional data and hence cannot establish causal effects. Furthermore, sample sizes are usually small and only a few interviewers conduct the surveys (Huddy et al. 1997). A major disadvantage of cross-sectional studies in this context is that interviewer effects are potentially confounded with area effects, as interviewers are typically allocated to specific geographic areas (Schnell and Kreuter 2005). To our knowledge, only two studies evaluating gender-of-interviewer effects related to gender issues address causality concerns by analyzing panel data, Zoch (2021) and Lipps and Lutz (2017). Zoch (2021) uses data from the National Educational Panel Study in Germany and analyzes how the interviewer’s gender affects self-reported gender ideologies. In line with previous research on gender issues, she finds that respondents report less traditional attitudes toward female interviewers. According to this study, there are, however, no significant gender-of-interviewer effects for an attitudinal item asking whether it’s the man’s job to earn money and the woman’s job to take care of the household and family. Housework hours or other time-use categories are not analyzed. The study by Lipps and Lutz (2017) is based on Swiss panel data 123 2982 M.Acht,M.Rebaudo from telephone interviews and evaluates gender-of-interviewer effects on a variety of different topics. Some of the analyzed items are related to factual household tasks (e.g., “in our household, it is mostly me who does the cleaning”) and the authors also analyze hours of housework but do not find significant gender-of-interviewer effects for any of these items. Because it has a different focus, the study does not analyze how results may differ by subgroups or other norm-related characteristics. It is therefore possible that potential gender-of-interviewer effects are hidden, for example, because in our study we only find effects for couples and not for singles. Furthermore, the direct interaction with an interviewer (as opposed to a telephone interview) is likely heightening the salience of norms which we argue to drive the effects. Our research question focuses on a factual survey question – time spent on housework – and hence differs from the majority of other studies about gender-of-interviewer effects that analyze attitudinal questions. Apart from Lipps and Lutz (2017) who do not find effects, there are no other studies evaluating gender-of-interviewer effects on housework hours. Importantly, housework hours are potentially very different from attitudinal questions about gender equality. In particular, the relationship between hours of housework and social norms is more subtle. While respondents may engage in impression management when answering attitudinal questions on gender equality for fear of offending the interviewer, this seems unlikely regarding the response about housework hours. Instead, we argue that the urge to fulfill a particular gender role is subconsciously affected by the interviewer’s presence. For example, respondents may become more or less aware of their own deviation from social norms depending on the interviewer’s gender. In line with self-deceptive reporting behavior this perception may subsequently influence the thought process of finding and formulating an answer, given that most people are uncertain about their exact housework hours. Our empirical analysis of different subgroups helps to disentangle the possible mechanisms for socially desirable reporting and supports this hypothesis of self-deceptive behavior. 3 Data We use data from the German Socio-Economic Panel (SOEP), a longitudinal survey of about 15,000 private households in Germany with annual interviews since 1984 (Goebel et al. 2019).1The majority of interviews in the SOEP are face-to-face interviews, which allows us to study the effect of the interviewer’s gender on individual responses. Notably, SOEP interviews are highly standardized. Interviewers are employed by the field institute Kantar, which prioritizes interviewer monitoring through ISO-certified processes that are regularly audited (Bohlender et al. 2020). For instance, at least 10% of annual interviews are checked, and each interviewer is monitored at least once a year. Interviewers are trained to ensure high data quality. They participate in annual training sessions conducted by the SOEP team at Kantar, along with representatives from DIW Berlin, where SOEP is based. These sessions prepare contact interviewers, who then train interviewers in the regions for which they are responsible. Contents of these trainings are scope, timing, and procedures of the 1Socio-Economic Panel (SOEP), data for years 1984-2018, version 35, 2020. 123 A gender gap in gender gaps… 2983 interviews, as well as any special features specific to each survey year (Goebel et al. 2019; Bohlender et al. 2020). Our main outcome of interest is the time spent on housework, which is measured using the following question (see the entire question in Fig.1in the Appendix): “What is a typical weekday like for you? How many hours per normal workday do you spend on housework (washing, cooking, cleaning)?”. This definition of housework aligns with the typical usage in the literature (e.g., Lippmann et al. 2020), specifically excluding other non-market activities like childcare or repair work and gardening, which are listed as separate activities in the same interview question. Routine housework is considered by both women and men as one of the least enjoyable activities (Gershuny 2013; Sullivan 2013), making it an interesting subject to study in the context of withincouple differences. Tasks like repair work and gardening, unlike routine housework, have been found to be more time flexible and more enjoyable (Coltrane 2000). As argued by Sullivan (2010), the norm concerning childcare has shifted since the 1970s, with time spent with children increasingly being viewed as leisure, particularly among more educated parents. Given that routine housework is most directly related to social norms, we restrict our focus on housework hours for the main specification. However, we also provide robustness checks for the other activities listed in the time-use question. Our analysis covers all years from 1985 to 2018, as information about the interviewers is available from 1985 onwards. We follow Lippmann et al. (2020) and Bertrand et al. (2015), and focus on working-age individuals between 18 and 65 years of age, since we can analyze the role of employment status and (relative) income for this group. The main sample consists of cohabiting couples (married or not married), although additional results for singles will be discussed in section 5. Importantly, in the SOEP data, all adults in the household answer the personal questionnaire, including the time-use reports, themselves. Thereby, within-couple gaps in housework reports can be analyzed. We exclude the few observations of partners who were interviewed by different interviewers.2We furthermore exclude more-generation-households and other non-typical household arrangements. Because the interviewer’s gender may differently affect homosexual individuals, we also exclude them from the analysis.3Our final sample consists of 18,965 couples that are observed for an average of around six years. Table 1presents descriptive statistics of the main sample. While women spend on average about three hours per weekday on housework duties, the average for men is 46min per day.4Of all individuals considered (employed and unemployed), the average working time of women is nearly half that of men, and incomes are consistently lower. Men in our sample are a bit older and on average have a slightly higher 2In the final sample this concerns only 68 observations. Couples are not excluded if partners are interviewed on different dates. In our sample 88% of couples are interviewed on the same day and around 95% within seven days after the first partner’s interview. Results are almost identical when excluding couples who were interviewed on different dates, see Table 11 in the Appendix. 3In our sample, there are only 112 different same-sex couples. 4In our sample women’s maximum housework time is 24h and men’s is 16h per day. When we assume 8h of sleep, housework estimates of more than 16h are likely overstated. However, only 32 observations report female housework hours to be larger than 16. Our results are robust when observations reporting high levels of housework, such as more than 16 h or even more than 6h, are excluded, see Table 12 in the Appendix. 123 2984 M.Acht,M.Rebaudo Table 1 Descriptive statistics Mean Standard deviation Min Max Woman’s housework time 3.00 1.95 0 24 Man’s housework time 0.76 0.93 0 16 Housework woman - housework man 2.24 2.23 −15 24 Paid work time woman 19.47 18.08 0 80 Paid work time man 36.69 18.69 0 80 Income woman 775.25 908.06 0 36184 Income man 2054.58 1725.40 0 98038 Woman’s age 42.58 10.72 18 65 Man’s age 45.33 10.77 18 65 Education level woman 4.92 2.33 0 9 Education level man 5.01 2.43 0 9 Married 0.88 0.33 0 1 No. of children age <6 0.30 0.59 0 6 No. of children age 6-12 0.43 0.73 0 5 No. of children age 13-16 0.23 0.50 0 4 Income HH 3447.94 2062.53 43 210084 East Germany 0.21 0.41 0 1 Observations 110,544 The table reports summary statistics of respondents’ characteristics based on the main sample (cohabiting couples of age 18 to 65). Housework time is hours per day. Paid work time is hours worked per week including overtime hours. Income is net monthly income and household income is net monthly income of all household members including non-labor income. Income is corrected for inflation with base year 2016. Education levels derive from the CASMIN classification. Married is a dummy variable equal to 1 for married couples. No. of children is the number of children living in the household for each of the age groups. East Germany is a dummy variable equal to 1 when the couple lives in East Germany at the time of the survey education level. 88% of the couples observed are married and around 21% live in East Germany at the time of the survey. In total, our sample consists of 1,841 different interviewers. Around 57% of all interviews are performed by male interviewers. Female and male interviewers differ in some characteristics that may affect interview responses, as shown in Table 2.Male interviewers on average have slightly less experience with SOEP, are older, more often married, more likely to have a higher education level, more often born in Germany, and more often of German mother tongue. Unfortunately, most questions about additional interviewer characteristics, except the interviewer’s age and experience, are only asked in specific years, so that including them as controls in the regression analyses results in a more selective sample. Therefore, we only include the interviewer’s gender and age for the main analysis and provide specifications with the additional characteristics as robustness checks in section 6.5Apart from the observed differences between 5Interviewer’s experience does not affect housework reports and the gender-of-interviewer effect is unaffected when additionally controlling for experience, see Table 13 in the Appendix. We do not include it as 123 A gender gap in gender gaps… 2991 the estimated coefficients, being in a couple relationship as opposed to being single, for women results in more housework hours, while for men it results in less. More interestingly for our purpose, the interaction effect between the Couple dummy and the interviewer’s gender for women is statistically significant and positive. Hence, the established result that the gender-of-interviewer effect is present for women in couple relationships but not for singles can also be found when studying transitions within individuals over time. As expected from the previous results there is no statistically significant interaction effect for men. The comparison of couples and singles helps to further disentangle the precise mechanisms behind socially desirable reporting, namely self-deception versus impression management. Above, we have argued that it seems unlikely for respondents to fear offending the interviewer, as housework hours are different from attitudinal items on gender equality. There are two additional explanations consistent with impression management as to why respondents may adjust their housework reports based on the gender of the interviewer. Firstly, findings on experimenter-gender-effects from other sciences suggest that there might be a general desire to impress one’s counterpart, especially in opposite-sex relationships (for heterosexual individuals) (Chapman et al. 2018). Individuals want to be perceived positively by the opposite sex, which can lead to self-affirming feedback or even the possibility of a romantic relationship. In relation to our results, this mechanism could indeed play a role if women were reporting more housework toward male interviewers, but it seems implausible that reporting less toward male interviewers would improve women’s perceived attractiveness as potential partners. Furthermore, if such a channel were in place, it should also be apparent for single women, since attractiveness on the marriage market is likely even more relevant for them. Hence, both the direction of the effect and the non-existence of effects for single women contradict such a channel. Secondly, and also in line with impression management theory, respondents may have a general desire to not being considered as unclean by other individuals when their own housework hours are perceived too low. If women expect higher cleanliness standards from other women as opposed to men, this channel could be in line with higher housework reports toward female interviewers. However, if those concerns were indeed driving the results, again one would expect similar effects for single women. As we do not find those, general expectations regarding cleanliness standards also do not seem to explain our results. As a consequence, subconscious self-deception instead of impression management seems to be the more relevant channel to explain the observed effect. By comparing couples versus singles, we compare gender-of-interviewer effects for subgroups which also differ with respect to their total time spent on housework. Therefore, we need to address the issue of absolute versus relative adjustments in the housework report. For instance, if the extent of gendered reporting depended on the baseline level of housework, the absolute gender-of-interviewer effect might only be small and insignificant for singles because they do less housework than women in couple relationships. Therefore, in Table 19 in the Appendix we analyze the genderof-interviewer effect in relative terms by using the logarithm of housework hours as the dependent variable. The results are qualitatively and quantitatively in line with the main 123 2992 M.Acht,M.Rebaudo Table 4 Housework time and interviewer gender - singles Singles Single-couple Women Men Women Men (1) (2) (3) (4) Female interviewer −0.046 0.018 −0.005 −0.005 (0.03) (0.03) (0.03) (0.02) Couple 0.190*** −0.396*** (0.03) (0.02) Female interviewer * couple 0.073** 0.023 (0.03) (0.03) Adjusted R20.101 0.054 0.171 0.088 Individuals 7564 4783 23668 21486 Observations 33912 20334 144456 130878 Dependent variable is female or male daily housework hours. The table reports the results of individual fixed effects estimations. Respondents and Interviewer controls as well as year and state fixed effects are included. Standard errors clustered at the individual level are in parentheses. Significance is denoted by *p<.10, ** p<.05, *** p<.01 specification. Importantly, even in relative terms there are no gender-of-interviewer effects for men or singles. To summarize, women in couple relationships report more housework hours toward female interviewers which we interpret as evidence for social norms affecting housework reports. Following the theory on social desirability, we expect gender-ofinterviewer effects to be larger when the deviation between what is perceived to be desirable and reality is larger. Moreover, gender-of-interviewer effects may depend on the internalization of traditional social norms. These hypotheses will be examined in the subsequent sections. 5.3 Gendered reporting and norm-related characteristics If the reporting of housework hours is driven by the desire to fulfill a certain role, the size of the effect may differ by different norm-related characteristics. Studies show that social norms can evolve over time (Fernández 2013) so that later-born generations possibly hold less traditional views about the role of women and men. With regard to our research question, it is therefore interesting to analyze whether the gender-ofinterviewer effect differs by respondent and interviewer cohorts. Table 5presents the results. We split respondents up into two groups by the threshold of 1960 as year of birth, which is roughly the sample mean. As columns 1 and 2 show, the gender-of-interviewer coefficient indeed appears to be smaller for the later born cohort of women, despite still being marginally significant. In column 3 we analyze how the effect differs by cohorts of interviewers. With respondent fixed-effects models we can observe variation in the interviewer’s cohort over time (but not in the respondents’ cohorts), so we split interviewers up into three instead of two groups. In 123 A gender gap in gender gaps… 2993 Table 5 Women’s housework time - respondent and interviewer cohorts Young cohort Old cohort All Young cohort Old cohort (1) (2) (3) (4) (5) Female interviewer 0.049* 0.117*** 0.103*** 0.078* 0.121** (0.03) (0.04) (0.03) (0.04) (0.05) [Reference: Interviewer born <1945] Interviewer born 1945-1959 −0.001 0.044 −0.063 (0.05) (0.05) (0.08) Interviewer born >=1960 −0.017 −0.022 0.040 (0.08) (0.09) (0.14) Female interviewer −0.015 −0.026 0.029 * Interviewer born 1945-1959 (0.05) (0.07) (0.08) Female interviewer −0.168** −0.150* −0.163 * Interviewer born >=1960 (0.08) (0.08) (0.16) Adjusted R20.156 0.184 0.167 0.156 0.184 Individuals 12027 6938 18965 12027 6938 Observations 63757 46787 110544 63757 46787 Dependent variable is female daily housework hours. The sample is restricted to cohabiting couples (age 18 to 65). Columns are based on respondent cohorts: Young cohort is defined as at least one person in the couple being born 1960 or later. The table reports the results of individual fixed effects estimations. Respondents and Interviewer controls as well as year and state fixed effects are included as in the main specification. Standard errors clustered at the individual level are in parentheses. Significance is denoted by *p<.10, ** p<.05, *** p<.01 123 2994 M.Acht,M.Rebaudo addition to the 1960-threshold that is comparable to respondents’ cohorts, we separately account for interviewers born before 1945 and interviewers born between 1945 and 1959. The gender-of-interviewer coefficient persists toward interviewers of the two older cohorts, which represent more than 90% of interviewers in our sample. However, there is a reverse effect for the youngest cohort of interviewers, resulting in an insignificant gender-of-interviewer coefficient for respondents being interviewed by these interviewers. When we compare the effects of the interviewer cohorts by the two respondent cohorts (columns 4 and 5), it appears that especially respondents from the younger cohort engage less in gendered reporting when interviewed by the youngest generation of interviewers, however, there may also be more transitions toward young interviewers in this group. The resulting gender-of-interviewer coefficient for the young respondent cohort toward young interviewers is not statistically different from zero. For male respondents, we still do not find any gender-of-interviewer effects when separately studying results by cohorts (see Table 20 in the Appendix). Summarizing, there is some evidence that gender-of-interviewer reporting is less present when respondents and more importantly interviewers are from younger cohorts and hence may hold less traditional views. This supports our hypothesis that gendered reporting is indeed related to social norms. We do not find that the interviewer’s age (independent of the birth cohort) is related to the gender-of-interviewer effect (see Table 21 in the Appendix). Furthermore, we analyze how results differ by education levels and by the presence of young children, as both may be related to the role respondents wish to fulfill. For example, mothers of small children may have a stronger urge to be perceived as “good” mothers by reporting more housework hours. Highly educated women, in contrast, may feel less pressure to conform to traditional roles. Interestingly, we find that the genderof-interviewer effect does not differ between respondents of low or high education and also not by the presence of children in the household (see Tables 22 and 23 in the Appendix). This further corroborates that gender-of-interviewer effects not only depend on respondents’ perceived social norms, but also on the deviation between the honest response and what is perceived to be desirable. By children and education status, such a deviation does not seem to differ much, despite potential differences regarding perceived social norms. 5.4 Gendered reporting and specialization in paid work If the reporting of housework hours is driven by the desire to fulfill a certain role, the size of the effect may differ by women’s employment status and income. We suppose that a woman’s employment status (in addition to her working hours) may affect her belief about which social role to fulfill: For example, full-time employed women – compared to other women – might feel less pressure to conform to the housewife role and hence feel less urge to (subconsciously) adjust housework reports. Then, the gender-of-interviewer effect that potentially captures the need to fulfill societal expectations, would be smaller for full-time employed women. Accordingly, a woman’s contribution to the couple’s income might also affect her belief about societal expectations regarding housework. In line with the previous hypothesis, it could be that 123 A gender gap in gender gaps… 2995 women with a higher contribution to household income feel less need to adjust housework reports to fulfill the housewife role because they are identifying themselves more as working women. In contrast, the “doing gender” literature would suggest that a woman earning more than her partner feels the need to conform to her traditional role by either doing more housework or, potentially, reporting more housework hours (Lippmann et al. 2020; Bertrand et al. 2015). According to this idea, women who outearn their male partners might show a stronger gender-of-interviewer effect because of the desire to conform to their social role. To test these hypotheses, we add women’s employment status and relative income as controls to the regression analyses. Results are presented in Table 6. For these regressions, we restrict the sample to dual-earner couples as the interpretation of differences in relative income and employment status is only meaningful for this group. In contrast to relative income, we define the full-time dummy in absolute terms for females (not in relative terms within the couple) because in our sample of dual-earner couples, 95% of men work full-time but only 43% of women do. Whether the woman is working full-time and her relative income are both included in the regressions as interaction terms with the interviewer’s gender and also used for subgroup analyses. Consistent with our first hypothesis, running separate regressions for full-time and not-full-time employed (part-time/irregularly employed) women (columns 1 and 2) shows that the gender-of-interviewer effect of the latter is almost twice as large as of the former. Hence, it seems to be the case that full-time employed women feel less need to adjust housework reports. Including women’s employment status as an interaction effect with the interviewer’s gender (column 3) suggests that full-time employment generally reduces housework hours, but the relationship between fulltime employment and the interviewer’s gender is not statistically significant. To study the role of relative income we define two variables, following Bertrand et al. (2015). Firstly, the variable RelativeIncomeWoman measures the woman’s share in the couple’s income, defined as IncomeWoman/(IncomeWoman+IncomeMan). Secondly, we define a dummy variable WomanEarnsMore coded one when the woman earns strictly more than her partner, and zero otherwise. When separately studying couples where the man earns more (or incomes are equal) and couples where the woman earns more (columns 4 and 5), we see that the gender-of-interviewer effect is smaller and not statistically significant in the latter group. This supports the first hypothesis that women with a higher contribution to the couple’s income feel less need to adjust their housework reports to fulfill a certain role. Furthermore, we run two additional regressions on the whole sample: In column 6, we control for women’s relative income and additionally include an interaction effect between the interviewer’s gender and a dummy variable indicating whether the woman earns more than the man.10 The relationships shown by Bertrand et al. (2015) and Lippmann et al. (2020)are also present when additionally controlling for the interviewer’s gender: Larger relative income decreases female housework hours, but there is a statistically significant 10 This specification is similar to the ones by Bertrand et al. (2015) and Lippmann et al. (2020)asthey control for the woman’s income share and additionally for a dummy variable indicating whether the woman earns more than the man. They do so in order to capture only the effect associated with the violation of the social norm that “a man should earn more than his wife” irrespective of specialization effects related to relative income. 123 2996 M.Acht,M.Rebaudo Table 6 Women’s housework time and interactions - dual-earner couples Employment Status Relative Income Full-time Not full-time All Male higher or equal Female higher All (1) (2) (3) (4) (5) (6) (7) Female interviewer 0.058** 0.107*** 0.096*** 0.088*** 0.042 0.079*** 0.130*** (0.03) (0.03) (0.03) (0.03) (0.04) (0.02) (0.04) Female interviewer −0.047 * Woman full-time (0.03) Woman full-time −0.121*** (0.03) Female interviewer −0.033 * Woman earns more (0.03) Female interviewer −0.167* * Relative income woman (0.10) Woman Earns More 0.127*** 0.113*** (0.02) (0.02) Relative income woman −1.148*** −1.077*** (0.10) (0.11) Adjusted R20.045 0.088 0.099 0.101 0.048 0.100 0.100 Individuals 7032 8890 13145 11861 3349 13145 13145 Observations 26695 34962 61657 51779 9878 61657 61657 Dependent variable is female daily housework hours. The sample is restricted to dual-earner couples (age 18 to 65). The table reports the results of individual fixed effects estimations. Respondents and Interviewer controls as well as year and state fixed effects are included. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 A gender gap in gender gaps… 2997 and positive coefficient for the WomanEarnsMoredummy, which represents the socalled doing gender effect. The interaction effect of the interviewer’s gender with the WomanEarnsMore dummy in this regression is not statistically significant. When the interviewer’s gender is interacted with relative income (instead of the dummy variable) (column 7), we observe a negative and marginally significant interaction effect. Thus, the higher the woman’s contribution to the couple’s income, the less they engage in gendered reporting. In both specifications, the overall gender-of-interviewer coefficient remains positive and statistically significant. These results combined with the previous ones suggest that women who are relatively more attached to paid work, feel less need to report more housework hours toward female interviewers. This is in line with the expectation that gender-of-interviewer effects will be larger when the deviation between what is expected to be desirable and reality is larger. While we foucs on dual-earner couples for these analyses, Table 24 in the Appendix also evaluates whether being employed (vs. not being employed) affects the results. As shown, the gender-of-interviewer effect does not differ by employment status at the extensive margin. As unemployed women typically do more housework than employed women, it is likely that they do not perceive a stronger deviation between their actual hours and their expected hours, compared to working women. We do not find evidence for the second hypothesis that women earning more than their partners react more strongly to the interviewer’s gender. Note, however, that these results neither confirm nor contradict the general idea of “doing gender” reporting, where women violating traditional social norms in the market sphere (by outearning their partners) report more housework hours to restore their gender identity. It could, for example, still be the case that these women report more housework hours, but do so similarly when confronted with female and male interviewers. Besides employment status and relative income, regional differences present another category of norm-proxies to test their relation to gender-of-interviewer effects: Following Alesina and Fuchs-Schündeln (2007), several studies exploit the natural experiment constituted by the 41-year division of Germany to study differences in social norms between Western and Eastern regions. Among studies evaluating gender attitudes and norms, results show that East Germans tend to hold more egalitarian gender-role attitudes than West Germans (Bauernschuster and Rainer 2012; Beblo and Görges 2018). Lippmann et al. (2020) analyze how the results presented by Bertrand et al. (2015) differ between West and East Germany and find substantial differences. Specifically, they show that “doing gender” behavior, i. e. the urge to compensate for violations of the “male breadwinner” norm (by manipulating labor market outcomes or by adjusting housework hours), is present for West Germans but not for East Germans. They explain the differences via the long-lasting effects of the more gender-equal East German institutions. Following the finding that East Germans hold less traditional views regarding the gender division of paid and unpaid work, it is interesting to evaluate whether and how the gender-of-interviewer effect might differ between West and East Germans.11 The results are presented in Table 7. Separately 11 East Germany here is defined as the current region of residence at the time of the survey and includes East Berlin but not West Berlin. Due to the geographical definition there is the possibility of within-respondent variation on the East dummy. Alternatively, East can be defined based on the biographical information about the residence region before 1989. Results are similar for this definition and available upon request. 123 2998 M.Acht,M.Rebaudo Table 7 Women’s housework time—West versus East Germany West East All (1) (2) (3) (4) (5) Female interviewer 0.077*** 0.058 0.078*** 0.105*** 0.140*** (0.03) (0.04) (0.03) (0.03) (0.05) East 0.374 0.252 0.094 (0.26) (0.27) (0.29) Female interviewer * East −0.011 −0.064 −0.077 (0.04) (0.06) (0.10) Woman full-time −0.166*** (0.03) Woman full-time * East 0.142*** (0.05) Female interviewer * woman full-time −0.070* (0.04) Female interviewer * woman full-time * East 0.111* (0.06) Relative income woman −1.235*** (0.13) 123 A gender gap in gender gaps… 2999 Table 7 continued West East All (1) (2) (3) (4) (5) Relative income woman * East 0.597*** (0.17) Female interviewer * relative income woman −0.204* (0.12) Female interviewer * East * relative income woman 0.211 (0.21) Adjusted R20.107 0.053 0.098 0.099 0.101 Individuals 10462 2748 13145 13145 13145 Observations 47792 13865 61657 61657 61657 Dependent variable is female daily housework hours. The sample is restricted to dual-earner couples (age 18 to 65). The table reports the results of individual fixed effects estimations. Respondents and Interviewer controls as well as year and state fixed effects are included. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 3000 M.Acht,M.Rebaudo studying East German and West German women as in columns 1 and 2, suggests that the gender-of-interviewer effect is smaller and not statistically significant for East German women. The interaction effect between the interviewer’s gender and East for the whole sample is not statistically significant. To gain further insight into differences between the groups, we evaluate how the relationship between full-time employment or relative income and the interviewer’s gender differs between East and West Germans. Specifically, we include a three-way interaction of the interviewer’s gender, the East dummy and women’s full-time employment in column 4 and women’s relative income in column 5. In both specifications, the general gender-of-interviewer effect remains statistically significant and there is no statistically significant interaction effect between East and FemaleInterviewer.As shown before in Table 6, full-time employed women generally do less housework, for East German women, however, there is a reverse positive interaction effect that is statistically significant. More interesting for our purpose are the interaction effects with the interviewer’s gender: While the interaction between the interviewer’s gender and women’s full-time employment is still negative and now even marginally significant, the three-way interaction with women’s full-time employment status is statistically significant and positive. Hence, full-time employed women engage less in gendered reporting, but for East Germans, there is a counteracting effect. The combined effect is not statistically significant (Coefficient 0.041, SE 0.049, p-value 0.403), suggesting that for East German women the gender-of-interviewer effect is not related to the employment status. Regarding the relation to women’s relative income, the results in column 5 imply that the negative effect of relative income on housework hours is smaller for East German women. Furthermore, similar to the results in Table 6,the gender-of-interviewer effect is lower when the woman’s contribution to the couple’s income rises. This negative effect is reversed for East German women, and the combined effect is again not statistically significant (Coefficient 0.007, SE 0.173, p-value 0.967). Hence, for East German women the gender-of-interviewer effect is also not significantly related to women’s relative income. Gender-of-interviewer effects and interaction terms with the several characteristics of specialization in paid work remain insignificant for male respondents (see Tables 25 and 26 in the Appendix). Summarizing, these results support the theory that gender-of-interviewer effects are less present for East German women compared to West German women. Greater attachment to paid work generally seems to reduce gender-of-interviewer effects, but for East German women this relationship is non-existent. This represents an interesting finding as traditional social norms and hence the urge to fulfill a housewife role are less prevalent for East Germans. Additionally, these results further support the conclusion that the established gender-of-interviewer effects are indeed driven by social norms. 5.5 Gendered reporting and cultural heritage Several studies have shown how social norms are related to cultural heritage (Alesina et al. 2013; Giuliano 2020). A recent paper by Blau et al. (2020) studies the role of source country characteristics, especially source country gender equality, in driving the division of non-market work among US immigrants. The major result of the paper 123 A gender gap in gender gaps… 3007 analysis (not shown here). Importantly, the gender-of-interviewer effect remains stable when only including couples who never have changes in the interview mode, which further alleviates concerns about the interview mode affecting the results. 7 Conclusion A growing body of literature seeks to understand how social norms and social roles affect behavior. In this field of research, the gendered division of housework hours has received increasing attention, not only by researchers but also in the public debate. Our results present evidence for the fact that social norms impact response behavior in interviews. We find a causal effect of the interviewer’s gender on women’s housework reporting: Women in couple relationships report more time spent on housework when they are interviewed by a woman instead of a man. We do not find any genderof-interviewer effects for male respondents, which leads to an observed interviewer gender gap in the housework gender gap. Therefore, estimates about time-use from face-to-face interviews should be interpreted with caution. We argue that the observed gender-of-interviewer effect is most likely driven by self-deceptive reporting behavior. Women internalize societal expectations more strongly when they interact with a female as opposed to a male interviewer, which subsequently affects their housework report. Stratifying the sample and interacting the interviewer’s gender with several norm-related characteristics support the hypothesis that the observed genderof-interviewer effects are driven by social norms. Note, while our results provide strong evidence for the existence of gendered reporting in time-use estimates, our findings do not imply that interviewer effects are the only way in which social norms affect the reporting in surveys. By comparing responses toward female and male interviewers, our approach captures only a part of potential adjustments in the reporting process. It is possible that social norms affect the reports beyond what is captured by the reaction to the interviewer’s gender. Importantly, it remains an open question whether even diary-based estimates of time-use are influenced by the social roles individuals try to fulfill. As there are no data on the “true” time-use values, it is impossible to compare interview reports to reality. Increasing the use of mixed-method approaches in survey design could be beneficial for testing differences between various modes concerning outcomes on a wide range of topics. While face-to-face interviews are known to have significant advantages compared to other interview modes (such as building trust on sensitive topics), our study implies that drawbacks may exist even for topics not previously considered as potentially affected by interviewer effects. More broadly, our findings offer a new perspective on the discussion about norms influencing behavior. While previous studies on socially desirable reporting have almost exclusively focused on normative views (e.g., about gender equality) or behaviors clearly linked to societal expectations (such as drug use or voting behavior), our study provides evidence that socially desirable reporting matters even for topics less clearly linked to social norms. Therefore, our understanding of the impact of norms on behavior may be limited by the prevalence of reporting biases for all kinds of outcomes. As a consequence, future research should take into account the conceivably 123 3008 M.Acht,M.Rebaudo strong effects of reporting behavior in domains not previously regarded as relevant to this issue. Appendix See Figs. 1,2,3and Tables 11,12,13,14,15,16,17,18,19,20,21,22,23,24,25, 26,27 and 28 in the appendix Fig. 1 Interview question - time-use. Time-use question of the personal interview questionnaire - SOEP 2018 123 A gender gap in gender gaps… 3009 Fig. 2 Share of female interviewers over time. The Figure is based on the main sample as defined in section 3.The bars represent the share of interviews with female interviewers over time Fig. 3 Share of female interviewers by interview month. The Figure is based on the main sample as defined in section 3.The bars represent the share of interviews with female interviewers by interview month. In less than 4 percent of cases interviews of partners were in a different month. For simplicity here only interviews with female respondents are shown 123 3010 M.Acht,M.Rebaudo Table 11 Housework time and interviewer gender - same day interview Women Men Gap (1) (2) (3) Female interviewer 0.080*** 0.012 0.068*** (0.02) (0.01) (0.03) Adjusted R20.165 0.100 0.169 Individuals 18149 18149 18149 Observations 97176 97176 97176 The table reports regression analyses for the main sample (cohabiting couples of age 18 to 65) but without couples who were interviewed at different dates. Dependent variable is daily housework hours of women, men or within-couple gap (Housework H our s W oman −Housework H ours Man). The table reports the results of individual fixed effects estimations. Additional controls as in the main specification. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 Table 12 Housework time and interviewer gender - restricting housework hours Housework hours <=16 <=12 <=8<=6 (1) (2) (3) (4) Dep. Var: Female housework hours Female interviewer 0.079*** 0.077*** 0.063*** 0.056*** (0.02) (0.02) (0.02) (0.02) Adjusted R20.170 0.169 0.170 0.161 Individuals 18933 18857 17937 16465 Observations 110142 109246 100555 87413 Dep. Var: Male housework hours Female interviewer 0.014 0.016 0.018 0.013 (0.01) (0.01) (0.01) (0.01) Adjusted R20.097 0.097 0.102 0.099 Individuals 18965 18961 18904 18825 Observations 110544 110509 109926 109170 The sample contains only cohabiting couples (age 18–65). Control variables as in the main specification. In all regressions we only keep observations who never exceed the respective threshold of housework hours. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 A gender gap in gender gaps… 3011 Table 13 Housework time and interviewer gender - interviewer experience Women Men Gap (1) (2) (3) Female interviewer 0.078*** 0.015 0.063*** (0.02) (0.01) (0.02) Interviewer experience 0.002 −0.001 0.003 (0.00) (0.00) (0.00) Adjusted R20.167 0.097 0.169 Individuals 18965 18965 18965 Observations 110544 110544 110544 The table reports regression analyses for the main sample (cohabiting couples of age 18 to 65). Dependent variable is daily housework hours of women, men or within-couple gap (Housework H ours W oman − Housewor k Hours Man). The table reports the results of individual (or couple) fixed effects estimations. Interviewer Experience measures the cumulative number of years interviewers have worked for SOEP. Additional controls as in the main specification. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 3012 M.Acht,M.Rebaudo Table 14 Respondent characteristics by interviewer gender Mean male interviewer Mean female interviewer Diff. SE diff. Income man 2029.94 2087.38 −57.434*** 10.484 Income woman 781.22 767.31 13.904* 5.518 Income HH 3437.45 3461.91 −24.467 12.534 Paid work time man 36.48 36.98 −0.493*** 0.114 Paid work time woman 19.54 19.38 0.157 0.110 Man’s age 45.45 45.18 0.273*** 0.065 Woman’s age 42.70 42.41 0.293*** 0.065 Education level man 5.01 5.02 −0.011 0.015 Education level woman 4.94 4.89 0.053*** 0.014 Married 0.88 0.88 0.002 0.002 No. of children age <60.30 0.30 −0.006 0.004 No. of children age 6-12 0.43 0.43 −0.002 0.004 No. of children age 13-16 0.23 0.23 0.003 0.003 Interv. day man 95.22 94.75 0.470 0.350 Interv. day woman 94.96 94.27 0.693* 0.350 Interview length man 35.10 36.40 −1.300*** 0.081 Interview length woman 34.21 35.39 −1.185*** 0.080 Observations 110,544 The table reports mean comparisons of respondents’ characteristics by interviewer gender based on the main sample (cohabiting couples of age 18 to 65). Paid work time is hours worked per week including overtime hours. Income is net monthly income and household income is net monthly income of all household members including non-labor income. Income is corrected for inflation with base year 2016. Education levels derive from the CASMIN classification. Married is a dummy variable equal to 1 for married couples. No. of children is the number of children living in the household for each of the age groups. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 A gender gap in gender gaps… 3013 Table 15 Housework time and interviewer gender - cross section Women Men Gap (1) (2) (3) Female interviewer 0.078*** 0.009 0.069*** (0.02) (0.01) (0.02) Respondent Characteristics Paid work time woman −0.037*** 0.007*** −0.044*** (0.00) (0.00) (0.00) Paid work time man 0.008*** −0.018*** 0.026*** (0.00) (0.00) (0.00) Income woman −0.000*** 0.000*** −0.000*** (0.00) (0.00) (0.00) Income man −0.000* −0.000*** 0.000** (0.00) (0.00) (0.00) Log. HH income −0.010 −0.022 0.012 (0.03) (0.02) (0.03) Married 0.169*** −0.069*** 0.238*** (0.02) (0.01) (0.02) No. of children age <6 0.352*** 0.090*** 0.262*** (0.02) (0.01) (0.02) No. of children age 6-12 0.325*** 0.040*** 0.286*** (0.01) (0.01) (0.01) No. of children age 13-16 0.325*** 0.026*** 0.299*** (0.01) (0.01) (0.02) Interviewer characteristics Age 0.001 0.000 0.001 (0.00) (0.00) (0.00) Interview length woman 0.002*** 0.001** 0.001 (0.00) (0.00) (0.00) Interview length man 0.003*** 0.001*** 0.002*** (0.00) (0.00) (0.00) Adjusted R20.415 0.189 0.406 Observations 110544 110544 110544 The table reports regression analyses for the main sample (cohabiting couples of age 18 to 65). Dependent variable is daily housework hours of women, men or within-couple gap (Housework H ours W oman − Housewor k Hours Man). The table reports the results in the cross section without controlling for individual fixed effects. Additional controls include age, age squared, education categories for women and men, respectively, interview mode, as well as year and state fixed effects as in the main specification. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 3014 M.Acht,M.Rebaudo Table 16 Housework time and interviewer gender - exogenous controls Women Men Gap (1) (2) (3) Female interviewer 0.082*** 0.010 0.072*** (0.02) (0.01) (0.03) Respondent characteristics Married 0.223*** −0.094*** 0.317*** (0.03) (0.02) (0.04) No. of children age <6 0.694*** 0.000 0.693*** (0.02) (0.01) (0.02) No. of children age 6–12 0.384*** 0.020** 0.364*** (0.02) (0.01) (0.02) No. of children age 13–16 0.227*** 0.023*** 0.203*** (0.02) (0.01) (0.02) Interviewer characteristics Age −0.001 0.001 −0.002 (0.00) (0.00) (0.00) Interview length woman −0.003*** 0.002*** −0.004*** (0.00) (0.00) (0.00) Interview length man 0.004*** −0.001*** 0.005*** (0.00) (0.00) (0.00) Adjusted R20.095 0.024 0.070 Individuals 18965 18965 18965 Observations 110544 110544 110544 The table reports regression analyses for the main sample (cohabiting couples of age 18 to 65). Dependent variable is daily housework hours of women, men or within-couple gap (Housework H ours W oman − Housewor k Hours Man). The table reports the results of individual fixed effects estimations. Additional controls include age, age squared, education categories for women and men, respectively, interview mode, as well as year and state fixed effects as in the main specification. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 A gender gap in gender gaps… 3015 Table 17 Total non-market work and interviewer gender Household Total chores non-market work Women Men Women Men (1) (2) (3) (4) Female interviewer 0.058* 0.024 0.208*** 0.040 (0.03) (0.02) (0.08) (0.05) Adjusted R20.130 0.146 0.306 0.135 Individuals 17549 17548 14459 14462 Observations 98965 98875 72060 72017 The table reports regression analyses for the main sample (cohabiting couples of age 18 to 65). Dependent variable is daily hours for women and men for the sum of household chores (housework, repairs/gardening, errands) in columns 1 and 2 and for the sum of all non-market activities (chores, childcare, elderly care) in columns 3 and 4. The table reports the results of individual fixed effects estimations. Additional controls include age, age squared, education categories for women and men respectively, interview mode, as well as year and state fixed effects. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 3016 M.Acht,M.Rebaudo Table 18 Gender-of-interviewer effects for other activities Childcare Repair/garden Elderly care Errands Education Sleep Work hours Exercise Leisure (1) (2) (3) (4) (5) (6) (7) (8) (9) Women Female interviewer 0.142*** 0.022* 0.002 0.006 −0.015 −0.006 0.031 0.015 0.053** (0.05) (0.01) (0.02) (0.01) (0.01) (0.03) (0.05) (0.02) (0.03) Adjusted R20.332 0.029 0.004 0.035 0.102 0.017 0.198 0.009 0.066 Individuals 18963 18959 14468 17554 18949 9369 18909 7675 18944 Observations 110521 110503 72124 98999 110399 31798 109138 23973 108560 Men Female interviewer 0.008 0.009 0.005 0.010 0.006 0.001 0.044 0.005 0.058** (0.02) (0.01) (0.01) (0.01) (0.01) (0.02) (0.04) (0.02) (0.03) Adjusted R20.101 0.057 0.003 0.059 0.099 0.017 0.402 0.014 0.108 Individuals 18965 18961 14469 17551 18957 9364 18900 7674 18938 Observations 110531 110413 72128 98993 110374 31797 109304 23971 108524 Dependent variable is daily hours for different activities. The sample is restricted to couples aged 18 to 65. The table reports the results of individual fixed effects estimations. Respondents and Interviewer controls as well as year and state fixed effects are included as in the main specification. For the regressions with work hours as dependent variable we exclude actual work hours as explanatory variable (it is not identical to the variable used in the main specification as here e.g. travel time is included but highly collinear). Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 A gender gap in gender gaps… 3023 Table 26 East interactions - male housework hours West East All (1) (2) (3) (4) (5) Female interviewer −0.000 −0.003 0.001 −0.009 −0.024 (0.02) (0.02) (0.02) (0.02) (0.03) East 0.275** 0.281** 0.338*** (0.12) (0.12) (0.13) Female interviewer * East 0.004 0.027 0.091 (0.03) (0.03) (0.06) Female interviewer * woman full-time 0.028 (0.02) Woman full-time 0.012 (0.02) Woman full-time * East −0.003 (0.03) Female interviewer * woman full-time * East −0.046 (0.04) Female interviewer * relative income woman 0.083 (0.08) Relative income woman 0.589*** (0.09) 123 3024 M.Acht,M.Rebaudo Table 26 continued West East All (1) (2) (3) (4) (5) Relative income woman * East −0.135 (0.11) Female interviewer * East * relative income woman −0.233 (0.15) Adjusted R20.037 0.012 0.031 0.031 0.033 Individuals 10462 2748 13145 13145 13145 Observations 47792 13865 61657 61657 61657 Dependent variable is male daily housework hours. The sample is restricted to dual-earner couples (age 18 to 65). The table reports the results of individual fixed effects estimations. Respondents and Interviewer controls as well as year and state fixed effects are included. Full-time and relative income are still defined for the female person in the couple for better interpretation. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 A gender gap in gender gaps… 3025 Table 27 Housework time and global gender gap index (GGI) - years spent in Germany Women Men Years in Years in Years in Years in Germany<20 Germany>=20 Germany<20 Germany>=20 (1) (2) (3) (4) (5) (6) Female interviewer 0.122** 0.169*** 0.048 0.030 0.013 0.034 (0.05) (0.06) (0.08) (0.02) (0.03) (0.03) Female interviewer * [GGI −GGI]−3.136*** −3.200*** −3.409** −0.133 0.038 -0.020 (1.02) (1.21) (1.58) (0.46) (0.63) (0.64) [GGI −GGI] 1.201 1.570 0.814 1.352*** 0.818 1.739*** (0.82) (1.02) (1.28) (0.47) (0.61) (0.63) Years in Germany 0.005 −0.004** (0.00) (0.00) Adjusted R20.350 0.374 0.326 0.165 0.175 0.164 Observations 12813 7298 5515 12826 6095 7130 Regression for the subsample of immigrant couples in which both partners originate from the same country of origin (age 18 to 65 in cohabiting couples). Dependent variable is daily housework hours of women, men, or within-couple gap. Couples 1-3 refer to female respondents, and columns 4-6 to male respondents. GGI is the Global Gender Gap Index. The table reports the results of OLS regressions. Respondents and Interviewer Controls as well as year and state fixed effects are included. Additional controls are source country fertility and GDP per capita. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 123 3026 M.Acht,M.Rebaudo Table 28 Housework time and interviewer gender - control for interview month Women Men Gap (1) (2) (3) Female interviewer 0.080*** 0.011 0.069*** (0.02) (0.01) (0.03) February 0.008 −0.024** 0.032 (0.02) (0.01) (0.02) March 0.037 −0.040*** 0.077*** (0.02) (0.01) (0.03) April 0.043* −0.047*** 0.090*** (0.02) (0.01) (0.03) May 0.034 −0.030* 0.064** (0.03) (0.02) (0.03) June 0.063** −0.027 0.091*** (0.03) (0.02) (0.03) July 0.082** −0.049*** 0.131*** (0.03) (0.02) (0.04) August 0.069* −0.030 0.098** (0.04) (0.02) (0.04) September 0.143*** −0.020 0.162*** (0.04) (0.02) (0.05) October 0.143** −0.059* 0.202*** (0.06) (0.03) (0.07) November 0.081 0.041 0.041 (0.12) (0.06) (0.14) December 0.207 0.157 0.050 (0.35) (0.11) (0.38) Adjusted R20.166 0.100 0.170 Individuals 18149 18149 18149 Observations 97176 97176 97176 The table reports regression analyses for the main sample (cohabiting couples of age 18 to 65) but without couples who were interviewed at different dates. Dependent variable is daily housework hours of women, men or within-couple gap (Housework H our s W oman −Housework H ours Man). The table reports the results of individual fixed effects estimations. Additional controls as in the main specification. Standard errors clustered at the individual level are in parentheses. Significance is denoted by * p<.10, ** p<.05, *** p<.01 Funding Open Access funding enabled and organized by Projekt DEAL. Data Availibility The SOEP data are not publicly available but researchers can sign a data distribution contract to access the data, https://doi.org/10.5684/soep-core.v35. The index of gender equality (GGI) comes from the World Economic Forum’s “The Global Gender Gap Report”, available at https://www3.weforum.org/docs/WEF_GenderGap_Report_2007.pdf. Total fertility and GDP data come from the World Bank, available at https://data.worldbank.org/indicator/ SP.DYN.TFRT.IN and https://data.worldbank.org/indicator/NY.GDP.PCAP.PP.KD. 123 A gender gap in gender gaps… 3027 Declarations Competing interests The authors have no financial or non-financial interests to disclose that are relevant to the content of this article. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. References Akerlof GA, Kranton RE (2000) Economics and identity. Q J Econ 115(3):715–753 Alesina A, Fuchs-Schündeln N (2007) Good-bye Lenin (or not?): the effect of communism on people’s preferences. Am Econ Rev 97(4):1507–1528 Alesina A, Giuliano P, Nunn N (2013) On the origins of gender roles: women and the plough. Q J Econ 128(2):469–530 Bauernschuster S, Rainer H (2012) Political regimes and the family: how sex-role attitudes continue to differ in reunified Germany. J Popul Econ 25(1):5–27 Beblo M, Görges L (2018) On the nature of nurture. The malleability of gender differences in work preferences. J Econ Behav Organ 151:19–41 Beblo M, Robledo JR (2008) The wage gap and the leisure gap for double-earner couples. J Popul Econ 21(2):281–304 Bertrand M, Kamenica E, Pan J (2015) Gender identity and relative income within households. Q J Econ 130(2):571–614 Bittman M, England P, Sayer L, Folbre N, Matheson G (2003) When does gender trump money? Bargaining and time in household work. Am J Sociol 109(1):186–214 Blau FD, Kahn LM, Comey M, Eng A, Meyerhofer P, Willén A (2020) Culture and gender allocation of tasks: source country characteristics and the division of non-market work among US immigrants. Rev Econ Househ 18(4):907–958 Bohlender A, Rathje M, Glemser A (2020) SOEP-Core - 2018. Report of Survey Methodology and Fieldwork Chapman CD, Benedict C, Schiöth HB (2018) Experimenter gender and replicability in science. Sci Adv 4(1):e1701427 Coltrane S (2000) Research on household labor: modeling and measuring the social embeddedness of routine family work. J Marriage Fam 62(4):1208–1233 Cooke LP (2007) Persistent policy effects on the division of domestic tasks in reunified Germany. J Marriage Fam 69(4):930–950 Deutsches Institut für Wirtschaftsforschung (2020). Socio-Economic Panel (SOEP), data for years 19842018: version 35, SOEP Fernández R (2013) Cultural change as learning: the evolution of female labor force participation over a century. Am Econ Rev 103(1):472–500 Flores-Macias F, Lawson C (2008) Effects of interviewer gender on survey responses: findings from a household survey in Mexico. Int J Pub Opin Res 20(1):100–110 Foster G, Stratton LS (2018) Do significant labor market events change who does the chores? Paid work, housework, and power in mixed-gender Australian households. J Popul Econ 31(2):483–519 Geist C (2009) One Germany, two worlds of housework? Examining employed single and partnered women in the decade after unification. J Comp Fam Stud 40(3):415–437 Gershuny J (2013) National utility: measuring the enjoyment of activities. Eur Sociol Rev 29(5):996–1009 Giuliano P (2020) Gender and culture. Oxf Rev Econ Policy 36(4):944–961 123 3028 M.Acht,M.Rebaudo Goebel J, Grabka MM, Liebig S, Kroh M, Richter D, Schröder C, Schupp J (2019) The German socioeconomic panel (SOEP). Jahrb Natl Stat 239(2):345–360 Hausmann R, Tyson L D, Zahidi S (2007) The Global Gender Gap Report 2007 Holtgraves T (2004) Social desirability and self-reports: testing models of socially desirable responding. Pers Soc Psychol Bull 30(2):161–172 Huddy L, Billig J, Bracciodieta J, Hoeffler L, Moynihan PJ, Pugliani P (1997) The effect of interviewer gender on the survey response. Polit Behav 19(3):197–220 Jessen J, Schweighofer-Kodritsch S, Weinhardt F, Berkes J (2024) Separate housework spheres. IZA DP No. 17134 Kan MY (2008) Measuring housework participation: the gap between “stylised” questionnaire estimates and diary-based estimates. Soc Indic Res 86(3):381–400 Kane EW, Macaulay LJ (1993) Interviewer gender and gender attitudes. Pub Opin Q 57(1):1–28 Kantar Public (2019). SOEP-Core - 2018. Personenfragebogen, Stichproben A-L3+N Krumpal I (2013) Determinants of social desirability bias in sensitive surveys: a literature review. Qual Quant 47(4):2025–2047 Kühne S (2018) From strangers to acquaintances? Interviewer continuity and socially desirable responses in panel surveys. Surv Res Methods 12(2):121–146 Leopold T, Skopek J (2015) Convergence or continuity? The gender gap in household labor after retirement. J Marriage Fam 77(4):819–832 Lippmann Q, Georgieff A, Senik C (2020) Undoing gender with institutions: lessons from the German division and reunification. Econ J 130(629):1445–1470 Lipps O, Lutz G (2017) Gender of interviewer effects in a multi-topic centralized CATI panel survey. Methods Data Anal 11(1):67–86 Liu M, Stainback K (2013) Interviewer gender effects on survey responses to marriage-related questions. Pub Opin Q 77(2):606–618 Lueptow LB, Moser SL, Pendleton BF (1990) Gender and response effects in telephone interviews about gender characteristics. Sex Roles 22(1–2):29–42 Murray-Close M, Heggeness M L (2019) Manning up and womaning down: how husbands and wives report earnings when she earns more. Opportunity and Inclusive Growth Institute Working Papers, 28 Paulhus DL (1991) Measurement and control of response bias. In: Robinson JP, Shaver PR, Wrightsman LS (eds) Measures of personality and social psychological attitudes. Academic Press, San Diego, pp 17–59 Press JE, Townsley E (1998) Wives’ and husbands’ housework reporting: gender class and social desirability. Gend Soc 12(2):188–218 Roth A, Slotwinski M (2020) Gender norms and income misreporting within households. ZEW Discussion Paper, No. 20-001 Schnell R, Kreuter F (2005) Separating interviewer and sampling-point effects. J Off Stat 21(3):389–410 Schober PS (2013) Maternal labor market return and domestic work after childbirth in Britain and Germany. Community Work Fam 16(3):307–326 Schober PS, Zoch G (2019) Change in the gender division of domestic work after mothers or fathers took leave: exploring alternative explanations. Eur Soc 21(1):158–180 Sullivan O (2010) Changing differences by educational attainment in fathers’ domestic labour and child care. Sociology 44(4):716–733 Sullivan O (2011) An end to gender display through the performance of housework? A review and reassessment of the quantitative literature using insights from the qualitative literature. J Fam Theory Rev 3(1):1–13 Sullivan O (2013) What do we learn about gender by analyzing housework separately from child care? Some considerations from time-use evidence. J Fam Theory Rev 5(2):72–84 Tourangeau R, Rips LJ, Rasinski KA (2000) The psychology of survey response. Cambridge University Press, Cambridge Tourangeau R, Yan T (2007) Sensitive questions in surveys. Psychol Bull 133(5):859–883 Warner RL (1986) Alternative strategies for measuring household division of labor. J Fam Issues 7(2):179– 195 West BT, Blom AG (2017) Explaining interviewer effects: a research synthesis. J Surv Stat Methodol 5(2):175–211 West C, Zimmerman DH (1987) Doing gender. Gend Soc 1(2):125–151 123 A gender gap in gender gaps… 3029 Zoch G (2021) Gender-of-interviewer effects in self-reported gender ideologies: evidence based on interviewer change in a panel survey. Int J Public Opin Res 33(3):626–636 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 123