methods, data, analyses | Vol. 16(1), 2022, pp. 3-32 DOI: 10.12758/mda.2021.

08

Gender and Survey Participation.

An Event History Analysis of the Gender
Effects of Survey Participation in a
Probability-based Multi-wave Panel Study
with a Sequential Mixed-mode Design

Rolf Becker
University of Bern

Abstract
In cross-sectional surveys, as well as in longitudinal panel studies, systematic gender dif-
ferences in survey participation are routinely observed. Since there has been little research
on this issue, this study seeks to reveal this association for web-based online surveys and
computer-assisted telephone interviews in the context of a sequential mixed-mode design
with a push-to-web method. Based on diverse versions of benefit–cost theories relating to
deliberative and heuristic decision-making, several hypotheses are deduced and then tested
by longitudinal data in the context of a multi-wave panel study on the educational and occu-
pational trajectories of juveniles. Employing event history data on the survey participation
of young panelists living in German-speaking cantons in Switzerland and matching them
with geographical data at the macro level and panel characteristics at the meso level, none
of the hypotheses is confirmed empirically. It is concluded that indirect measures of an
individual’s perceptions of a situation, and of the benefits and costs as well as the process
and mechanisms of the decision relating to survey participation, are insufficient to explain
this gender difference. Direct tests of these theoretical approaches are needed in future.

Keywords: Gender; survey participation; nonresponse; event history analysis; societal

environment; panel study; web-based online survey; sequential mixed-mode
design; push-to-web method
© The Author(s) 2021. This is an Open Access article distributed under the terms of the
Creative Commons Attribution 3.0 License. Any further distribution of this work must
maintain attribution to the author(s) and the title of the work, journal citation and DOI.
4 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

Against the background of decreasing response rates in modern societies with a

high level of prosperity, the number of empirical studies on survey participation
and nonresponse in the social sciences is increasing (e.g. Leeper, 2019; Beullens
et al., 2018; Dutwin & Lavrakas, 2016; Keusch, 2015; Tourangeau & Plewes, 2013;
Brüggen et al., 2011; Stoop et al., 2010; Groves & Peytcheva, 2008; Groves, 2006;
Groves et al., 2001; de Heer, 1999; Smith, 1995; Goyder & Leiper, 1985; Steeh,
1981). One of the main findings is that the decrease in response rates is generally
observed for cross-sectional surveys, while the participation rate within longitudi-
nal studies, such as panel studies, remains high (Becker et al., 2019; Brick & Wil-
liams, 2013; Schoeni et al., 2013: 84–85). Among these studies, a constant gender
effect in survey participation and nonresponse is observed as a social phenom-
enon (Slauson-Blevins & Johnson, 2016: 428; Keusch, 2015: 186; Busby & Yoshida,
2013; Dykema et al., 2012; Laguilles et al., 2011; Stoop et al., 2010: 10, 20; Couper
et al., 2008: 260; Marcus & Schütz, 2005; Patrick et al., 2013; Porter & Whitcomb,
2005; Kwak & Radler, 2002: 259; Curtin et al., 2000: 419; Singer et al., 2000: 180;
Green, 1996: 176; Dalecki et al., 1988: 54). Particularly for mail or web surveys,
it is frequently found that women are more likely to respond than men (Green,
1996: 176; Becker & Glauser, 2018). Furthermore, women seem to be more likely
than men to respond promptly after the invitation to take part in an online survey
(Becker, 2021; Becker et al., 2019). Finally, over the last several years, it has been
found for different survey modes and survey topics that the gender effect on the
rate of survey participation remains clear, even though response rates are declining
overall (Slauson-Blevins & Johnson, 2016). However, it is still not known whether
the gender difference in participation rates changes across surveys in a multi-wave
panel.

Acknowledgements
The data for the first seven waves of the panel study are available as Scientific Use Files
at FORS in Lausanne and can be found in the online catalogue under the reference
number 10773 (https://forsbase.unil.ch/project/study-public-overview/15802/0/). The
data of Wave 8 will be available in 2021. These scientific use files include the paradata
of the fieldwork periods.
For helpful comments on an earlier version, I wish to thank the anonymous reviewers
and the mda editor Sabine Häder. The author is responsible for the remaining short-
comings.
Funding
The DAB panel study is substantially financed by the State Secretariat for Education,
Research and Innovation (SERI). The interpretations and conclusions are those of the
author and do not necessarily represent the views of the SERI.

Direct correspondence to
Rolf Becker, University of Bern, Department of Sociology of Education,
Fabrikstrasse 8, CH–3012 Bern, Switzerland
E-mail: rolf.becker@edu.unibe.ch
Becker: Gender and Survey Participation 5

In this respect, Green (1996: 176) states there is too little research on gender
as it relates to surveys to reach a conclusion. However, there are several “ad-hoc
explanations” of the effects of gender on response rates. Slauson-Blevins and John-
son (2016), for example, emphasize that the lower inclination of men to take part
in scientific surveys might cause the decrease in their rate of response. “Gender
differences in survey participation are partially attributable to the difficulty of con-
tacting male participants rather than outright refusals to participate (…). Yet while
survey researchers, often conclude that gender differences exist, there has been lit-
tle effort to conceptually understand this difference” (Slauson-Blevins & Johnson,
2016: 428). Another explanation that seems plausible attributes gender disparities
in survey response to differences in socialization regarding norms around helping,
or differences in susceptibility to social influence. External circumstances, such as
access to the Internet in a prosperous country like Switzerland, do not contribute
to the explanation of these gender disparities since there is no “digital divide” in
Internet use across the genders (BFS, 2021).
Focusing on self-administered survey modes, such as an online questionnaire
or administered computer-assisted telephone interview (CATI), the question is still
unsolved regarding why gender has been found to play a significant role in survey
participation and response to questionnaires or interviewers, with women respond-
ing in greater proportion than men (Porter & Whitcomb, 2005). Likewise, it is
unclear why female and male online panelists are motivated differently (Slauson-
Blevins & Johnson, 2016; Göritz & Stieger, 2009). Are there gender-specific moti-
vations, resources, and circumstances that drive male and female invitees to par-
ticipate in different ways? Against the theoretical background of a diverse variety
of rational action theories that take heuristic decision-making process into account,
the main question asked in this empirical contribution is as follows: Are gender
differences in survey participation a fundamental phenomenon or are they epiphe-
nomenal to other factors, such as social origin and class-related socialization in
terms of educational level and achievement? In other words: are gender differences
a singular phenomenon observed for cross-sectional surveys or in early waves in a
multi-wave panel study? Do gender disparities in surveys disappear when we con-
trol for a number of covariates, which correlates with the propensity toward survey
participation?
To answer these research questions, longitudinal data on survey participation
are needed. The optimal type of longitudinal data would be the observation of a
target person’s survey participation across their life course, including time-variant
information on their resources, circumstances and preferences. Since such data
combining individual information on target persons with survey paradata are rare,
the measurement of survey participation in a panel study provides a suboptimal
surrogate. Therefore, data on the survey participation of female and male panel-
ists collected since 2012 in an event history design are utilized in this contribu-
6 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

tion. This type of data, collected in the context of a multi-wave panel study on
educational and occupational trajectories of juveniles born around 1997 and living
in German-speaking cantons in Switzerland, makes it possible to analyze gender
differences in overall survey nonresponse, the development of these gender differ-
ences during the fieldwork, and changes in them across surveys for a single target
sample familiar with Internet and mobile devices.
In the remainder of this contribution, the next section outlines the theoretical
background, as well as the hypotheses to be tested. The following section com-
prises a description of the data, design, statistical procedures, and the variables.
After then, the empirical findings are presented. The final section gives a summary
and conclusion.

Theoretical Background
Despite the fact that there is no theoretical vacuum regarding survey methodolo-
gies, Singer (2011: 379) concluded that, although various theories of survey par-
ticipation exist, we know comparatively little about why individuals are willing or
are not willing to participate, and about how they decide to take part in (or refuse
to take part in) a scientific survey. Although different versions of rational action
theories – such as social exchange theory (Dillman, 2000), the theory of subjective
utility (Becker & Mehlkop, 2011), leverage-salience theory (Groves et al., 2000) or
the social-psychological approach on habitual-heuristic action (Groves et al., 1992)
– all assume that survey participation is based on a deliberative assessment of the
benefits and costs of survey participation, or on an automatic-spontaneous deci-
sion, there is no comparative test of these approaches in theoretical and empirical
respects. However, such a systematic test is needed to confirm Singer’s (2011: 388)
conclusion that the general benefit-cost theory of survey participation can be seen
as a synthesis of principles derived from these other theories (Goyder, Boyer, &
Martinelli, 2008). Thus, it is unclear whether gender differences in survey partici-
pation can be explained by a target person’s reasoned judgment that the benefits of
acting outweigh the costs, or by an almost instantaneous cognitive procedure with
the help of heuristics (Singer, 2011: 381).
According to these approaches, survey participation According is a function of subjectivesurvey particip
to these approaches,
perceived costs and benefits of survey participation,𝐵𝐵asof well as the participation,
survey subjective expectedas well as
According to these approaches, survey participation 𝑠𝑠𝑠𝑠 is a function of the subj
subjecti
probability of successful realization of perceived benefits
benefits 𝑝𝑝: 𝑓𝑓�𝑠𝑠𝑠𝑠� � 𝑠𝑠 ∙ 𝐵𝐵
𝐵𝐵 of survey participation, as well as the subjective expected probability of su – .
𝐶𝐶. The decision
The decision regarding
benefitsparticipation
𝑝𝑝: 𝑓𝑓�𝑠𝑠𝑠𝑠� �or𝑠𝑠refusal
∙ 𝐵𝐵 – 𝐶𝐶. assessment
is based
The on of subjective
a subjective
decision expected
assessment
regarding utilitiesorSEU
participation re
of subjective expected utilitiesofSEU(.)
assessment of different
subjective expected If 𝑆𝑆𝑆𝑆𝑆𝑆�𝑠𝑠𝑠𝑠�
alternatives � 𝑠𝑠
da, other
utilities SEU(.) ∙ 𝐵𝐵
than� 𝐶𝐶
�� survey
��of different � 𝑆𝑆𝑆𝑆𝑆𝑆�𝑑𝑑𝑑𝑑�
�� alternatives da, �o
participation sp. If 𝑆𝑆𝑆𝑆𝑆𝑆�𝑠𝑠𝑠𝑠� � 𝑠𝑠�� ∙ 𝐵𝐵�� � 𝐶𝐶�� � part in the survey
𝑆𝑆𝑆𝑆𝑆𝑆�𝑑𝑑𝑑𝑑� � 𝑠𝑠�� ∙ 𝐵𝐵�� � 𝐶𝐶�� , it is likely an el
part
is likely an eligible in the survey
individual will indeed take part in the survey.
However, what gender-specific costs or, in particular, gender-specific bene-
fits of survey participation might there be? According to Singer (2011), a potential
Becker: Gender and Survey Participation 7

respondent’s decisions depend mainly on benefits, not on costs or perceptions of

risk or harm. Why should female target persons systematically perceive increased
benefits resulting from survey response than their male counterparts? A plausible
answer might be that it is the gender-specific expectations of success probabil-
ity that result in gender differences regarding the benefits of survey participation.
These expectations might be based on an individual’s skills – such as literacy or
computer skills – and the related confidence in their own abilities (e.g. persis-
tence; decisiveness; internal or external control beliefs). In theoretical respects, this
assumption is based on empirical evidence that girls and young women have better
educational achievement, higher language proficiency and more advanced literacy
and educational attainment than boys and male adolescents (DiPrete & Buchmann,
2013; Beck et al., 2010; Becker & Müller, 2011; Buchmann et al., 2008). However,
it has to be taken into account that there is a stark correlation between educational
success, educational attainment and social origin among the genders. These educa-
tional advantages in favor of female target persons indicate that cognitive burden,
uncertainty in interview situations and insufficient language ability and proficiency
are much lower for women compared to men. Since the transaction costs of survey
response are relatively lower for women, they are more likely to take part in a sci-
entific survey than male potential respondents.
Hypothesis 1: Controlling for social origin, educational attainment and
achievements (indicating language proficiency and language ability, as well as
educational success and motivation, mostly in favor of women) are positively cor-
related with survey participation. Due to the advantage of female panelists in edu-
cational success and achievements, the gender effect becomes insignificant when
these dimensions are taken into account.
Furthermore, the correlations between gender, educational attainment, social
origin and the rural-urban divide in educational opportunity are evident (Glauser
& Becker, 2016; Sixt, 2013). In sum, according to Green (1996), education, intel-
ligence and achievement, as well as socioeconomic status and living in rural areas,
were found to correlate positively with survey response rate (Becker, 2021; Groves
& Couper, 1998; Dalecki et al., 1988: 54).
Hypothesis 2: By controlling additionally for regional opportunity structures
in terms of a potential respondent’s place of residence in a rural or urban area, the
gender difference in survey participation diminishes in multivariate estimations.
Success in educational attainment is correlated with favorite educational
returns in an individual’s working life. Although women profited from educational
expansion (e.g. Becker & Mayer, 2019; Becker & Müller, 2011), recent research
has argued that the opportunity costs of survey participation are higher for men, as
women are more likely to stay at home (Stoop et al., 2010: 20). Women are therefore
more likely to be reachable and ready to take part in a survey. But this line of rea-
soning might not be valid for self-administered web-based online interviews, since
8 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

for these invitees can decide themselves whether and when they will start complet-
ing the questionnaire – e.g. after the working day, at the weekend or at another
point in time suitable for them. This might be also true for CATI due to widespread
availability of mobile devices. For juveniles in particular, it is confirmed that most
possess a smartphone instead of a fixed phone line (BFS, 2021).
If there are gender-based preferences for survey modes – i.e. that men are
more likely to prefer computer-assisted web-based interviews (CAWI) (due to their
technical affinity) and women the CATI mode (due to their language abilities) – it
could be assumed that there is no gender difference of response in surveys with a
sequential mixed-mode design. Since it is often observed, even for online surveys,
that women tend to respond earlier than men after survey launch (e.g. Göritz, 2014;
Göritz & Stieger, 2009), a sequential mixed-mode design offering CAWI and CATI
modes could have the potential to compensate for the gender-based likelihood of
participation in different survey modes in the long run of the fieldwork period.
Hypothesis 3: In a sequential mixed-mode design, the gender difference in
survey participation diminishes across the running fieldwork period and the offered
survey modes.
Furthermore, Green (1996) argues that gender differences may exist in sur-
vey response due to differences in (primary) socialization regarding differences
in susceptibility to social influence or helping norms. These aspects correspond
with findings by Porst and von Briel (1995: 11) that, besides personal and situative
aspects, women are more likely to respond to surveys due to altruism (Porst & von
Biel, 1995: 15). In line with theoretical arguments on gender-based secondary and
tertiary socialization across the life course, it seems that girls and women display
a different social character than boys and men (e.g. Grunow, 2013). For example,
compared to their male counterpart, they are more likely to have learned to carry
out a task – such as the request of another person or completing a questionnaire –
in an autonomous, precise and persistent way (e.g. Quenzel & Hurrelmann, 2010).
According to Green (1996: 181), women are therefore more communicative and
interested in sharing opinions with others.
Hypothesis 4: Gender differences in survey participation disappear in multi-
variate estimations when controlling for personality traits and individual beliefs,
indicating at least some facets of gender-specific socialization.
For male youth and adults oriented toward traditional “masculinity norms”
or the “male breadwinner model”, it has been observed they are less interested in
tasks such as reading and writing, as well as in constructive communication with
other persons and authorities (e.g. Hadjar, 2011). On the one hand, this again means
that personality traits (such as persistence and decisiveness or internal and external
control beliefs) could help explain the gender differences in survey participation
(e.g. Porst & von Briel, 1995). One the other hand, it seems that the gender differ-
ences result from the low propensity toward survey participation observed for male
Becker: Gender and Survey Participation 9

target persons oriented toward the traditional gender stereotypes and gendered life
courses.
Hypothesis 5: Gender differences in survey participation are statistically dis-
solved by taking a panelist’s orientation toward gender roles into account, as well as
their personality traits and other individual skills.

Data, Design, Variables and Statistical Procedures

Data set
The empirical analysis is based on longitudinal data of a probability-based multi-
wave panel study about the determinants of educational choice and training oppor-
tunities (for details, see Becker et al., 2020). This project started in 2012. The
last survey was realized in May/June 2020. Data and paradata were collected in
a sequential mixed-mode design with a push-to-web method (see also: Kreuter,
2013). The first mode was an online survey, followed by a CATI and, in a selected
number of surveys, a paper-and-pencil interview (PAPI) by mail survey. The initial
target population comprised eighth-graders in the 2011/12 school year (born around
1997), who were enrolled in regular classes in public schools in German-speaking
cantons of Switzerland. The panel data are based on a random and 10 per cent
stratified gross sample of 296 school classes, out of a total universe of 3,045 classes.
A disproportional sampling of school classes from different school types, as well
as a proportional sampling of school classes regarding share of migrants within
schools, was applied. At the school level, a simple random sample of school classes
was chosen. The initial probability sampling was based on data obtained from the
Swiss Federal Statistical Office (FSO) (for details, see Glauser, 2015).
In the first three waves, the contacted panelists (n ≈ 3,800) were interviewed
in the context of their school class. After leaving the compulsory school, the panel-
ists were pursued individually after the fourth wave. Each of the eligible and con-
tactable panelists was invited for the surveys, even when they had skipped a wave.
To improve the response rate, the panelists received unconditionally prepaid mate-
rial incentives or cash in hand (Becker et al., 2019). Across the panel waves, the
overall response rate was constant at about 80 per cent (Table 1). The response rate
for online survey increased from 46 per cent in Wave 4 to 76 per cent in Wave 8,
while the response rate for the CATI decreased from 38 per cent to 5 percent.
The proportion of women among the invitees was rather constant, at 50 per
cent in Waves 4 and 5, 51 per cent in Wave 6 and 52 per cent in the Waves 7 and
8 at the start of survey launch. At the start of the risk time for CATI, about 47
per cent of the nonrespondents, i.e. invitees who had not taken part in the CAWI
before, were female in Wave 4. Their share decreased to 43 per cent in Wave 6 and
remained constant for the recent waves.
10 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

Table 1 Samples and response in the DAB panel

Wave 4 Wave 5 Wave 6 Wave 7 Wave 8

Oct–Nov Jun–Aug May–Jun May–Jun May–Jun
2014 2016 2017 2018 2020

Sample size
Contactable individuals 2,655 2,799 2,712 2,488 2,492
Type of survey
Online survey yes yes yes yes yes
CATI survey yes yes yes yes yes
PAPI survey no no yes yes no
Incentive voucher voucher pen money money
Realized interviews
Individuals 2,235 2,228 2,053 1,957 2,016
of whom: online 1,227 1,329 1,375 1,645 1,884
CATI 1,008 899 597 287 132
PAPI 0 0 81 25 0
Response rate in %
Contactable individuals 84% 80% 76% 79% 81%
Online 46% 47% 51% 66% 76%
CATI 38% 32% 22% 12% 5%
PAPI – – 3% 1% –

Source: DAB (own calculation)

For the analysis of gender effects on survey participation, the empirical analy-
sis focuses on the online and CATI modes only since the number of participants
in the PAPI mode was rather low (106 cases out of 13,145 target persons across six
panel waves, i.e. a response rate of 3% in Wave 6 and 1% in Wave 7). The observa-
tion window was standardized to 52 days for methodological reasons, such as com-
parability between waves, low number of participants after seven weeks of field-
work and right-censored data due to survey nonresponse. In the case of both survey
modes, non- and under-coverage was rather low for this sample. About 93 per cent
of the Swiss population has access to the Internet and they mostly use this medium
every day of the week. Each of the young interviewees in this panel study had daily
access to the Internet or possessed a telephone or other mobile device (BFS, 2021).
In total, 13,145 complete cases were available for analyzing gender-specific
patterns of participation in at least one of the five surveys. Since time stamps – col-
lected automatically by the online survey software Unipark or by the CATI soft-
ware – indicated exact time references for the invitation sent by email or SMS and
Becker: Gender and Survey Participation 11

the start of a panelist’s response, it was possible to calculate the exact duration of
episodes since survey launch on a daily basis (Becker, 2021; Durrant et al., 2013).
For the analysis of participation in the CATI by nonrespondents in the initial survey
mode, the waiting time was calculated on a daily basis by the difference between
the invitation to the CATI mode and the data of the telephone interview. For the
invitees who did not take part at all until the end of the fieldwork period, i.e. the
censored cases, the waiting time was 52 days. The number of skipped events was
negligible.
The distribution of these waiting times from invitation until an individual
started the survey participation as a stochastic event was analyzed using the tech-
niques and procedures of event history analysis (Blossfeld, Rohwer, & Schneider,
2019). This means that episodes of survey participation are the units to be analyzed.
In this respect, it was possible simultaneously to analyze an individual’s intention to
participate in the survey and the timing of when they did so. At the aggregate level,
the development of the response rate was observed across different points in time
during the field period.
For our purpose, this data set provides additional advantages due to the survey
design. Multiple waves, for example, ensure that a constant gender effect on sur-
vey participation is not random. These waves are associated with different prepaid
incentives, but with the same features of survey management; it is therefore possi-
ble to reveal if a gender effect depends on the type of an incentive by controlling for
cover letters (including the incentive), digital invitation and reminders. The sample
consisted of members belonged to a single birth cohort. Therefore, their survey
participation did not depend on the age of the panelists. The survey topic – their
own educational and occupational trajectory – was a general one and not related
directly to gender. The number of items on gender-related issues, such as gender-
based socialization or gender-based inequalities, was rather limited because the
primary task of the panel was the reconstruction of their educational trajectories
and careers in the labor market. Each of the target persons was involved in training
or employment so different states and time constraints in this regard did not matter
for survey participation. In respect of sponsorship and authority, it was emphasized
in the advance letter that the project was in receipt of a grant from a governmental
agency and was conducted by the same researchers at a Swiss university. Further-
more, for the sampling, 106 regions – characterized by a certain spatial homogene-
ity and reflecting small partially cross-cantonal labor market areas with functional
orientation toward centered and peripheral opportunities and living standards, in
addition to urbanicity, population density and a lack of social cohesion (Couper &
Groves, 1996: 174) – were considered (Glauser & Becker, 2016: 20). This allowed
for an analysis of the rural-urban divide in gender-specific survey participation
in terms of Internet access and living conditions. In sum, the data allowed for a
dynamic longitudinal analysis by considering the macro, meso and micro level –
12 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

i.e. social environmental attributes, survey characteristics and respondent attributes

– at the same time.

Dependent and independent variables

The dependent variable was the time-dependent likelihood of participation in the
CAWI. In general, the participation rate was defined by the ratio of contacted target
persons who completed the questionnaire or the telephone interview (RR1 accord-
ing to AAPOR, 2016: 61; Tourangeau & Plewes, 2013: 11; Bethlehem, 2009: 213;
Singer, 2006: 637). This variable was coded in the following way: “1” for participa-
tion in online survey, “2” for participation in the CATI mode and “0” for nonre-
sponse or incomplete response. Across the five panel waves, a maximum of 0.1 per
cent of respondents canceled their completion of the questionnaire in a survey.
The main independent variable was a panelist’s gender. It was considered
as a dummy variable, with men as the reference category. Another covariate was
the individual’s educational level, indicated by the school type in which they were
enrolled in their compulsory schooling. The school type was a proxy for the indi-
vidual’s appreciation of the utility of social-scientific research and information-
gathering activities associated with their education (Groves & Couper, 1998: 128).
The following school types were distinguished along their basic, extended and
advanced requirements: low, intermediate and academic level. The target person’s
achievement was measured by the (z-standardized) grade point average (GPA) in
German taught at school; this covariate indicated their cognitive resources and lan-
guage proficiency (Wenz, Al Baghal & Gaia, 2021). Using a dummy variable, it
was controlled for that German was the first language, indicating the target person’s
language ability (Kleiner, Lipps & Ferrez, 2015). By the way, this indicator mea-
sured the impact of migration background – net of German mother tongue, edu-
cational level and social origin – on survey response (Kalter, Granato & Kristen,
2007). Social origin was taken into account as a proxy for the socioeconomic con-
ditions in which the target persons grew up, including welfare, integration and envi-
ronment (Groves & Couper, 1998: 30). This was indicated by the well-established
class scheme suggested by Erikson and Goldthorpe (1992). Personal characteristics
– such as persistence, internal and external control belief and decisiveness – were
controlled for (Marcus & Schütz, 2015; Saßenroth, 2013). They were extracted
Becker: Gender and Survey Participation 13

from a number of items by factor analysis.1 The gender role models for women and
men were considered – after their extraction by factor analysis – for the indication
of gender-specific socialization.2
Another covariate was the current panel wave, indicating the effect of differ-
ent prepaid incentives, such as vouchers (worth 10 Swiss Francs), a ballpoint pen
(worth 2 Swiss Francs) or cash (10 Swiss Francs), as well as the panelist’s experi-
ences with this panel. The opportunity structure of the region in which the panel-
ists live was measured by macro data on regional levels delivered by the Swiss FSO.
In order to reduce complexity and to control for the high correlation of regional
contextual characteristics, factor scores were extracted from these data (for details:
Glauser & Becker, 2016). The 106 regions in the German-speaking cantons were
characterized by a certain spatial homogeneity and reflected the principle of small,
partially cross-cantonal labor market areas with functional orientation toward cen-
tered and peripheral opportunities and living standards, in addition to urbanicity,
population density and a lack of social cohesion.

Statistical procedures
Overcoming the limits of comparative-static estimations of survey response, the
techniques and statistical procedures of event history analysis were utilized (Bloss-

1 They were measured in the first and second waves. Persistence was measured by the
respondent’s agreement with the following five statements: “I do not like unfinished
business”; “If I decide to accomplish something, I manage to see it through”; “I com-
plete whatever I start”; “Even if I encounter difficulties, I persistently continue”; and
“I even keep at a painstaking task until I have carried it through”. The control beliefs
were measured by six items indicating the respondent’s internal and external locus of
control, as suggested by Jakoby and Jacob (1999): “I like to take on responsibility”;
“Making my own decisions instead of relying on fate has proved to be good for me”;
“In the case of problems and resistance, I generally find ways and means to assert
myself”; “Success depends on luck, not on performance”; and “I feel like I have little
influence over what happens to me”. Decisiveness was based on a question about the
respondent’s decision certainty: “Life is full of decisions that need to be taken. Which
of the six statements apply to you?” The wordings of these statements were: “I am re-
ally unsure as to what I should decide and often waver back and forth”; “Others unsettle
me in my decision”; “After making a decision, I have great doubts as to whether I really
made the right decision”; “It is very hard for me to decide because there are so many
possibilities”; and “When I make a decision, I stick to it”. For each of these items, the
agreement itself consisted of a scale of discrete values from 1 for “I strongly disagree”
to 5 for “I strongly agree”. In order to reduce complexity and to avoid multicollinearity,
three factors – persistence, control beliefs and decisiveness – were extracted by factor
analysis (Table A-1 in the Appendix).
2 The items of gender role stereotyping are measured in the third wave. Separately for the
genders, the respondents were asked their subjective view of whether it was interesting
for women or men to be employed, to earn much income, to be successful in their ca-
reer, to have children, to take care of the household and to be responsible for childcare.
The possible answers ranged from “1” for complete rejection to “5” for complete agree-
ment (Table A-2 in the Appendix)
 14 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

feld et al., 2019). In this contribution, the aim is to model the likelihood of survey
g the limits of comparative-static estimations of survey response, the techniques and statistical
participation – that is, the hazard rate – as a stochastic and time-variant function of
of event history
tive-static estimations analysis were utilized
ofresources,
survey response, (Blossfeld et al., 2019).
theoftechniques
techniques In this contribution, the aim is to
andsocietal
statistical
ive-static estimations
individual of survey response,
the settings the the survey and
and statistical factors. This rate r(t)function
is
likelihood
were utilized
s were of survey
utilizeddefined participation
(Blossfeld et al., – that
2019). is,
In the
this hazard rate
contribution, – asthe a stochastic
aim is to and time-variant
(Blossfeld as the et al., 2019).
marginal In
value this contribution, the aim is to
al resources,
cipation
ipation thatthe
––that is,the
is, settings
thehazard
hazard of rate
the
rate–survey
–as and of
asaastochastic
stochastic
the conditional
societal andfactors.
and
probability
This
time-variant
time-variant rate ��𝑡𝑡�
function
function
of issuch an event
defined occur-
as the marginal
e conditional ring – namely
probability of the start
such an of completing
event occurring –the questionnaire
namely the start in
of acompleting
web-basedthe online sur-
questionnaire
of the
of the survey
survey and and societal
societal factors.
factors. ThisThis rate
rate��𝑡𝑡���𝑡𝑡� isis defined
defined as as the
the marginal
marginal
ased
fsuch
such online
ananeventsurvey
event vey –– in
in the
occurring
occurring the time
––time
namely
namely interval
interval
thestart
the �𝑡𝑡, 𝑡𝑡of
start �completing
of giventhat
�𝑡𝑡�, given
completing that this event
this eventhas
thequestionnaire
the questionnaire has notnotoccurred
occurred before
before time
d et
etime al., 2019:
timeinterval 29).
time
interval�𝑡𝑡, Using
t this
(Blossfeld statistical
giventhat
�𝑡𝑡�,given
�𝑡𝑡,𝑡𝑡𝑡𝑡���𝑡𝑡�, et al.,
thatthis procedure,
2019:
thisevent 29).
eventhas it
Using
hasnot is possible
this
notoccurred to
statistical
occurredbefore reveal impacts
procedure,
beforetimetime of
it 𝑥𝑥
is for the
possible probable
to
of survey
hisstatistical
his participation
reveal impacts
statisticalprocedure,
procedure, as the
ititisispossibleevent
of x for to
possible the𝑦𝑦 to be
probable
toreveal
reveal investigated:
occurrence
impacts
impacts of𝑥𝑥𝑥𝑥for
of 𝑋𝑋
ofthe
for →
survey
the
� participation as the event 𝑡𝑡𝑡 by
Pr�𝑌𝑌
probable
probable �� �, whereby 𝑡𝑡 �
asiming of
the event
s the events
event 𝑦𝑦𝑦𝑦yto into
totobe
be account
beinvestigated:
investigated: 𝑋𝑋
investigated: 𝑋𝑋�� → Pr�𝑌𝑌�����,�,, whereby
→ Pr�𝑌𝑌 whereby
whereby 𝑡𝑡𝑡𝑡 �� 𝑡𝑡𝑡𝑡𝑡𝑡 by bytaking the timing of
unt
nt events into account.
Due to the sequential mixed-mode design, specialties of the timing of events
had to be considered for the bivariate and multivariate analyses. In the sequential
mixed-mode design of the DAB panel study, access to the online mode was pos-
sible for each of the invitees during the complete field period. Nonrespondents were
asked to take part in the CATI mode about two weeks after survey launch. There
was then a competing risk of taking part in one of the two offered modes, which
are mutually exclusive during an overlapping risk period. A competing risk is an
event – such as participation in one of the two survey modes – that either hinders
the occurrence of the primary event of interest (e.g. participation in the online sur-
vey instead of CATI) or that modifies the chance that this event (e.g. participation
in CATI) will occur (Noordzij et al., 2013: 2670). When eligible panelists prefer
one mode or another, the unchosen mode cannot be realized at another point in
time due to censoring. Panelists who have not started completing the questionnaire
have the “chance” to take part in the CATI or online mode at a point in time that is
convenient for them.
In the case of competing risks, the traditional survival analysis is inadequate
for methodological reasons. Therefore, the cumulative incidence competing risk
method was used to describe panelist participation patterns across the field period.
For example, the cause-specific cumulative incidence function (CIF), which is the
probability of survey participation before the end of field period, was estimated to
reveal the risk of choosing one of the competing survey modes (Lambert, 2017).
The CIF describes the incidence of the occurrence of an event while taking com-
peting risks into account (Austin & Fine, 2017: 4293).
Furthermore, parametric regression procedures were used to estimate the
impact of independent variables on the likelihood of interesting events. For this pur-
pose, the subdistribution hazards approach by Fine and Gray (1999) was seen as
the most appropriate method to use for analyzing competing risks (see also Schus-
ter et al., 2020; Noordzij et al., 2013). By taking competing risks into account, the
coefficients estimated by the stcrreg module implemented in the statistical package
Stata could be used to compute the cumulative incidence of participation in one
of the survey modes and to depict the hazards in a CIF plot (Austin & Fine, 2017).
Becker: Gender and Survey Participation 15

Finally, non-parametrical procedures were utilized to describe gender differ-

ences in survey participation. On the other hand, the parametrical procedures were
used in the sense of residual analysis. This means it was necessary to test whether
the gender difference in survey participation “disappeared” when controlling for
theoretically based variables. In this way, it was possible to decide if the gender dif-
ference was fundamental or based on other factors correlated with gender, such as
educational level, language proficiency or socialization.

Empirical Results
Description of gender-specific participation rates
If one measures both the timing and the quantity of participation in the online sur-
vey across several panel waves, the gender disparity becomes obvious. In the left-
hand panel in Figure 1, it is apparent for each of the panel waves that the likelihood
of survey participation in terms of cumulative incidences was significantly higher
for female panelists than for their male counterparts. In particular, the differences
in speed and rates increased in the initial stage after the survey launch. After about
two weeks, the development of these incidences was rather similar for both genders.
For Waves 4 and 5, the same patterns different for genders were observed for
the cumulative incidence of participation in the consecutive CATI mode since it
was offered to the nonresponding panelists (right-hand panel in Figure 1). While
there were no gender differences in taking part in this mode for Wave 6, men were
more likely to respond to CATI than women in the both most recent Waves 7 and 8.
It is evident for the CAWI mode that women started to complete the question-
naire earlier than male panelists. Across each of the waves, after 10 days, 50 per
cent of the female panelists had taken part in the online survey, while after 15 days
half of the male risk sample had completed the online questionnaire.3 The situation
was different for the CATI mode. For women, the median value for the CATI mode
was 15 days since this survey mode was offered to the panelists; this value was 16
days for their male counterparts. For this survey mode, except for Waves 5 and 7,
there were no systematic gender differences in participation. In sum, while 62 per
cent of the female panelists took part in the online survey and 39 per cent of female
nonrespondents, who do not responded in the CAWI mode yet, took part in the

3 For the CAWI mode, each of the tests – such as the Wilcoxon-Breslan-Gehan test,
sensitive at the beginning of the process time, or the Generalized Savage Log-rank test,
stressing increasing differences at the end of the process time – provided significant
differences between the compared units, such as gender and waves (Blossfeld et al.,
2019: 83). The null hypothesis that the timing and quantity of survey participation do
not differ across genders and waves must therefore be rejected for the initial survey
mode.
16 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

Figure 1 Gender disparities in survey participation across panel waves

Becker: Gender and Survey Participation 17

Figure 2 Gender disparities in survey participation across survey modes (esti-

mated by competing risk model)

CATI mode, 51 per cent of the male panelists completed the online questionnaire
and 40 per cent of them who did not respond in the initial survey mode took part in
the CATI. The overall participation rate across the fieldwork period of 52 days was
82 per cent for women and 76 per cent for male panelists.
Finally, this gender disparity in survey participation was confirmed for each of
the waves by multivariate analysis. For each of the waves, so-called β-coefficients
for gender estimated by competing risk analyses are depicted in Figure 2.4 In sum,
this finding again provides evidence for gender disparities of participation different
for the survey modes offered in a sequential mixed-mode design with a push-to-
mode strategy. They are significant for each of the waves in the CAWI mode. For
the CATI mode, it was observed that young women were more likely to participate
at the survey, while there was a reverse gender disparity in Wave 7. Overall, this
finding does not confirm Hypothesis 3 proposing that gender disparities in survey
participations diminish in a sequential mixed-mode design across the surveys and

4 The whiskers in the plot present the 95-% confidence interval of the coefficients. If they
cross the vertical zero line, these effects are insignificant.
18 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

panel waves. Finally, it became obvious that an extension of the fieldwork period
did not always result in decreasing gender differences in survey participation. In
spite of three digital reminders in the CAWI mode and a sequence of reminders in
the CATI mode after three call attempts, the participation rate declined completely
to zero after four or five weeks.

Parametric analysis of gender-related participation rates

Utilizing a competing risk model, stark and statistically significant gender dif-
ferences in survey participation were confirmed again for the initial online mode
(Table 2). On average, by controlling for panel wave and regional opportunity
structure, the inclination of female panelists to take part was () 53 per cent higher
than men’s (Model 1.1).
For the CATI mode, however, there were no significant gender differences in
survey response (Model 1.2). Among the nonrespondents to whom the CATI mode
was offered, there was no gender disparity in the timing and quantity of survey
response for the entire number of panel waves. Furthermore, as described above,
the differentiation of survey participation again made it obvious that participa-
tion in the CAWI increased across the panel waves, while the propensity toward
response in the CATI mode decreased for recent panel waves. Finally, the effect
of regional opportunity structure was only significant for the initial survey mode,
where the response rates were lower in urban areas compared to the rural context.
Living in urban areas resulted in a lower rate and speed of survey participation
after the survey launch. This result does not confirm Hypothesis 2, since gender
disparities remain constant.
Additionally, there was an impact of social origin on survey participation
(Models 2.1-2.2). The selectivity of survey participation in terms of social ori-
gin was characterized by the fact that panelists from the middle and upper social
classes had a greater inclination to complete the questionnaire than children of less
skilled and unskilled blue-collar workers. Working class children were more likely
to postpone their response and not take part in the CAWI or CATI than panelists
from the other social classes. In contrast to Hypothesis 1, the gender differences in
survey response were not explained by the social origin of panelists.
Panelists with a high educational level were more likely to take part in the
online surveys than individuals enrolled in lower secondary schools with basic
requirements. High language proficiency and ability in German language was
correlated with starting early to complete the online questionnaire, while these
achievements and skills were insignificant for participation in the CATI. By con-
trolling for social origin, educational level and language, there were still gender
disparities in survey participation in the initial survey mode. Therefore, Hypothesis
1 is not in line with these findings.
Table 2 Gender and participation in different panel waves of the DAB panel study

Survey mode CAWI CATI CAWI CATI CAWI CATI

Models 1.1 1.2 2.1 2.2 3.1 3.2
Gender
Female 0.356 –0.046 0.273 –0.052 0.290 –0.038
(0.022)*** (0.038) (0.023)*** (0.038) (0.023)*** (0.039)
Waves (Ref.: Wave 4)
Wave 5 0.034 0.035 0.045 0.063 0.046 0.064
(0.039) (0.046) (0.040) (0.046) (0.040) (0.046)
Wave 6 0.109 –0.396 0.124 –0.359 0.125 –0.359
(0.039)** (0.053)*** (0.039)** (0.054)*** (0.039)** (0.054)***
Wave 7 0.502 –0.848 0.505 –0.822 0.507 –0.822
Becker: Gender and Survey Participation

(0.037)*** (0.066)*** (0.037)*** (0.067)*** (0.037)*** (0.067)***

Wave 8 0.757 –1.476 0.772 –1.442 0.774 –1.441
(0.036)*** (0.092)*** (0.036)*** (0.092)*** (0.036)*** (0.092)***
Macro factor
Regional opportunity structure –0.037 –0.016 –0.046 –0.016 –0.045 –0.014
(0.011)** (0.018) (0.012)*** (0.019) (0.012)*** (0.019)
Social origin (Ref.: Upper service class)
Lower service class 0.012 –0.067 0.007 –0.068
(0.039) (0.067) (0.039) (0.067)
Routine non-manual employees –0.011 –0.029 –0.008 –0.034
(0.037) (0.062) (0.037) (0.062)
Farmers, small proprietors –0.066 –0.002 –0.061 –0.002
(0.054) (0.092) (0.054) (0.092)
Foremen, skilled manual workers –0.153 –0.152 –0.143 –0.146
(0.042)*** (0.067)* (0.042)*** (0.067)*
19
 20

Survey mode CAWI CATI CAWI CATI CAWI CATI

Models 1.1 1.2 2.1 2.2 3.1 3.2
Semi- and unskilled manual workers –0.139 –0.027 –0.137 –0.027
(0.057)* (0.094) (0.057)* (0.093)
Missing value –0.140 –0.224 –0.138 –0.226
(0.045)** (0.075)** (0.045)** (0.075)**
School type (Ref.: Basic requirements)
Extended requirements 0.508 0.146 0.494 0.132
(0.032)*** (0.045)** (0.032)*** (0.045)**
Advanced requirements 0.915 0.240 0.896 0.217
(0.038)*** (0.066)*** (0.038)*** (0.066)**
Missing value 0.382 0.002 0.373 –0.007
(0.042)*** (0.067) (0.042)*** (0.067)
Language
Language proficiency (GPA) 0.128 –0.032 0.124 –0.037
(0.013)*** (0.021) (0.013)*** (0.021)
Language ability (German vs. others) 0.225 0.060 0.228 0.052
(0.034)*** (0.052) (0.034)*** (0.052)
Personality traits
Persistence 0.054 –0.015 0.057 –0.010
(0.013)*** (0.020) (0.013)*** (0.020)
Control belief 0.035 0.015 0.036 0.017
(0.012)** (0.020) (0.012)** (0.020)
Decisiveness 0.049 0.046 0.045 0.042
(0.012)*** (0.019)* (0.012)*** (0.019)*
Gender role orientation
Female role model –0.086 –0.045
(0.014)*** (0.023)*
methods, data, analyses | Vol. 16(1), 2022, pp. 3-32
 Survey mode CAWI CATI CAWI CATI CAWI CATI
Models 1.1 1.2 2.1 2.2 3.1 3.2
Male role model 0.054 –0.013
(0.015)*** (0.022)

Number of episodes 13,145 6,898 13,145 6,898 13,145 6,898

Number of events 7,460 2,744 7,460 2,744 7,460 2,744
Number of competing risks 2,923 1,392 2,923 1,392 2,923 1,392
Number of censored cases 2,762 2,762 2,762 2,762 2,762 2,762
Wald chi2 (d.f.) 1,016.2 (6) 446.7 (6) 2178.6 (20) 510.6 (20) 2222.8 (22) 516.6 (22)
* p<0.05; ** p<0.01; *** p<0.001; β-coefficients, estimated by competing risk model (in brackets: robust standard error; clustered for individual
units).
Becker: Gender and Survey Participation

Source: DAB (own calculations)

21
22 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

Figure 3 Impact of gender role orientation on survey participation (estimated

by competing risk model)

The effects of personal traits were also straightforward. Panelists with high
persistence, distinct primary or internal control belief and pronounced decisive-
ness were more likely to take part in the CAWI than individuals with external
or secondary control belief, or individuals who were indifferent or characterized
by remissness (Model 2.1). It was also found that panelists who postponed their
response were more likely to take part in the CATI provided they had pronounced
decisiveness (Model 2.2). However, since the gender difference was still significant,
Hypothesis 4 – stressing that personality traits explain the gender differences in
survey participation – is not supported empirically.
While a panelist’s orientation toward a traditional female role model made
them less likely to respond to an invitation to the CAWI (Model 3.1) and to the CATI
(Model 3.2), it was obvious that panelists who agreed with the “male breadwinner
model” were more likely to be motivated in (early) survey participation (Model 3.1).
However, if interaction effects of gender and gender role orientation are taken into
account, by controlling for the same variables as in the models 3.1 and 3.2, there
was no significant effect of them on survey response (Figure 3). This was true for
the online mode (left-hand panel) as well as for the CATI mode (right-hand panel).
Overall, these interaction effects on response were very small and did not dissolve
the gender differences in survey participation at all. Therefore, Hypothesis 5 is not
Becker: Gender and Survey Participation 23

confirmed empirically, proposing that the effects of gender role orientation on sur-
vey response explain the gender differences in survey participation.
Finally, this issue was also true for the interactions effect of gender and each
of the other covariates considered in model estimations, such as panel experience,
social origin, educational level, language ability and proficiency, and regional
opportunity structure. Each was insignificant; therefore, they are not reported or
discussed in detail.

Discussion
In the dynamic analysis of the likelihood and timing of survey participation, the
empirically evident gender differences could not be discounted by taking theoreti-
cally proposed processes and mechanisms into account, at least indirectly. Even if
factors at the macro level of societal environment, at the meso level of survey char-
acteristics and at the micro level of an interviewee’s resources, abilities and beliefs
were considered in the event history analysis, the gender effect on the timing and
quantity of survey participation remained significant. None of the different hypoth-
eses considering gender-based processes and mechanisms at each of the analytical
levels was confirmed empirically. It seems that there are unobserved heterogeneities
in gender-specific preferences and circumstances, and those perceptions of benefits
and costs of responses in surveys of a multi-wave panel are not taken into account
in a way that would support the assumptions of the theory of subjective expected
utility and the heuristic logic of habitual action regarding scientific surveys.

Summary and Conclusions

The manifest aim of this empirical analysis has been to contribute to an evidence-
based explanation of systematic gender disparities in survey participation. The
latent aim is to relaunch this issue as a matter of interest in the research on sur-
vey methods. Regarding survey methodology, this research issue is still notoriously
under-investigated in contemporary survey methodology (Becker, 2021: 20; Green,
1996). Thus, the question to be answered by this analysis was why we continu-
ously observe differences between the genders regarding survey participation and,
in particular, in its timing and quantity. Why are female target persons more likely
to take part in social-scientific surveys than men?
Utilizing event history data on the likelihood of young panelists participat-
ing in surveys within a single-cohort and multi-wave panel study conducted in
German-speaking cantons of Switzerland, the analysis has attempted to explain
the gender differences in survey participation by hypotheses deduced from an
advanced version of reasoned action theory and heuristic decision-making (Singer,
2011). According to Green (1996), it is assumed that, among other influences, the
24 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

gender difference is mainly based on gender-specific abilities, skills and achieve-

ments, which can be indicated by an invited potential respondent’s language profi-
ciencies and abilities, as well as by their educational success and attainment. Since
girls and women have become advantaged in this respect due to educational expan-
sion, it seemed plausible that a male target person’s lower propensity toward survey
participation might be correlated with their educational level and skills, resulting
in gender disparities of participation. Even when social origin – providing a direct
influence on an individual’s educational achievement and attainment – was taken
into account, the gender differences remained constant in each of the surveys. In
the panel with a sequential mixed-mode design and a push-to-web-method, the
gender differences were obvious for the initial online mode. However, even when
other influences (such as personality traits, agreement with traditional gender roles
or living in a rural or urban region) were taken into account, the gender disparities
remained unsolved.
This result could be based on some limitations of this contribution. First of
all, there is no elaborated theory explaining gender differences in survey partici-
pation. Ad-hoc arguments dominate a coherent explanation. Ideally, this theory
should be a special case of a rational action approach. Second, the data provided
less information on the mechanisms relevant for explaining response in general
(e.g. benefit-cost calculation) that should be integrated into the statistical models.
There was also a lack of information regarding different circumstances for genders
that were essential for assessing the likelihood of survey participation. Third, the
target population was limited to juveniles of a single birth cohort living in a small
area in a small country. However, it could be argued that an explanation of gender-
based survey response should be universal.
While none of the different hypotheses was confirmed empirically (and have
not been confirmed in previous studies), and since the residual analysis conducted
in the context of a multi-wave panel study on the educational and occupational
trajectories of juveniles born around 1997 was not successful at all, the search for
an empirically tested answer on the association between gender and survey par-
ticipation must continue. Future studies may more profitably address the incre-
mental effects of gender by directly measuring individual preferences, expecta-
tions and motivations, as well as perceived benefit-cost balance and everyday life.
The social mechanisms emphasized in the wide variety of rational action theories
and approaches to heuristic decision-making must also be observed directly with
systematic reference to gender. As a by-product of such an endeavor, the different
theories attempting to explain survey participation per se could be tested. Such a
comparative test of theories on survey participation is overdue.
Becker: Gender and Survey Participation 25

References
AAPOR (The American Association for Public Opinion Research) (2016). Standard Defi-
nitions: Final Dispositions of Case Codes and Outcome Rates for Surveys. AAPOR
(9th edition).
Austin, P.C., & Fine, J.P. (2017). Practical recommendations for reporting Fine‐Gray Model
Analyses for competing risk data. Statistics in Medicine, 36(27), 4391–4400.
Beck, M., Jäpel, F., & Becker, R. (2010). Determinanten des Bildungserfolgs von Migranten
im Schweizer Bildungssystem. In G. Quenzel & K. Hurrelmann (Eds.), Bildungsverlie-
rer – Neue Ungleichheiten. (pp. 313–337). Wiesbaden: VS.
Becker, R., Möser, S., & Glauser, D. (2019), Cash vs. vouchers vs. gifts in web surveys of
a mature panel study – Main effects in a long-term incentives experiment across three
panel waves. Social Science Research, 81, 221–234.
Becker, R., & Glauser, D. (2018). Are prepaid monetary incentives sufficient for reducing
panel attrition and optimizing the response rate? An experiment in the context of a
multi-wave panel with a sequential mixed-mode design. Bulletin of Sociological Meth-
odology, 137, 74–95.
Becker, R. (2021). The Effects of a Special Sequential Mixed-Mode Design, and Reminders,
on Panellists’ Participation in a Probability-Based Panel Study. Quality and Quantity,
55, 1–26.
Becker, R., & Mehlkop, G. (2011). Effects of prepaid monetary incentives on mail survey
response rates and on self-reporting about delinquency – Empirical findings. Bulletin
of Sociological Methodology, 109, 5–25.
Becker, R., & Müller, W. (2011). Bildungsungleichheiten nach Geschlecht und Herkunft im
Wandel. In A. Hadjar (Ed.), Geschlechtsspezifische Bildungsungleichheiten (pp. 55–
75). Wiesbaden: VS Verlag für Sozialwissenschaften.
Glauser, D., & Becker, R. (2016). VET or general education? Effects of regional opportunity
structures on educational attainment in German‑speaking Switzerland. Educational
Research in Vocational Education and Training, 8, 1-25
Becker, R., & Mayer, K. U. (2019). Societal Change and Educational Trajectories of Women
and Men Born between 1919 and 1986 in (West) Germany. European Sociological Re-
view, 35(2), 147–168.
Becker, R., Glauser, D., & Möser, S. (2020). Determinants of educational choice and vo-
cational training opportunities in Switzerland – Empirical analyses with longitudinal
data from the DAB panel study. In N. McElvany, H. G. Holtappels, F. Lauermann, A.
Edele, & A. Ohle-Peters (Eds.), Against the Odds – (In)Equity in Education and Edu-
cational Systems (pp. 125–143). Münster: Waxmann.
Bethlehem, J. (2009). Applied Survey Methods. Hoboken, N.J.: Wiley.
Beullens, K., Loosveldt, G. Vandenplas, C., & Stoop, I. (2018). Response rates in the Eu-
ropean Social Survey: increasing, decreasing, or a matter of fieldwork efforts? Survey
Methods: Insights from the Field. Retrieved 4 July 2020, from the Survey Methods
website: https://surveyinsights.org/?p=9673.
BFS (Bundesamt für Statistik – Federal Office for Statistics) (2021). Internetnutzung. Re-
trieved on May 5, 2021, from the website of the Swiss Federal Statistical Office: https://
www.bfs.admin.ch/bfs/de/home/statistiken/kultur-medien-informationsgesellschaft-
sport/informationsgesellschaft/gesamtindikatoren/haushalte-bevoelkerung/internet-
nutzung.html.
26 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

Blossfeld, H.-P-, Rohwer, R., & Schneider, T. (2019). Event History Analysis with Stata.
London, New York: Routledge.
Brick, M.J., & Williams, D. (2013). Explaining rising nonresponse rates in cross-sectional
surveys. AAPSS (The ANNALS of the American Academy of Political and Social Sci-
ence), 645(1), 36–59.
Brüggen, E., Wetzels, M., de Ruyter, K., & Schillewaert, N. (2011). Individual differences
in motivation to participate in online panels. The effect on response rate and response
quality perception. International Journal of Market Research, 53(3), 369–390.
Buchmann, C., DiPrete, T.A., & McDaniel, A. (2008). Gender inequalities in education. An-
nual Review of Sociology, 34, 319–337.
Busby D. M., & Yoshida, K. (2013). Challenges with online research for couples and fami-
lies: evaluating nonrespondents and the differential impact of incentives. Journal of
Child and Family Studies, 24(2), 505–513.
Couper, M.P., & Groves, R.M. (1996). Social environmental impacts on survey cooperation.
Quality & Quantity, 30(May), 173–186.
Couper, M.P., Kapteyn, A., Schonlau, M., & Winter, J. (2011). Noncoverage and nonre-
sponse in an Internet survey. Social Science Research, 36(1), 131–148.
Couper, M.P., Singer, E., Conrad, F.G., & Groves, R.M. (2008). Risk of disclosure, percep-
tions of risk, and concerns about privacy and confidentiality as factors in survey par-
ticipation. Journal of Official Statistics, 24(2), 255–275.
Curtin, R., Presser, S., & Singer, E. (2000). The effects of response rate changes on the In-
dex of Consumer Sentiment. Public Opinion Quarterly, 64(4), 413–428.
Dalecki, M.G., Ilvento, T.W., & Moore, D.E. (1988). The effects of multi-wave mailings on
the external validity of mail surveys. Journal of the Community Development Society,
19(1), 51–70.
de Heer, W. (1999). International response trends: results of an international survey. Journal
of Official Statistics, 15(2), 129–142.
Dillman, D.A. (2000). Mail and Internet Surveys. The Tailored Design Method. New York:
Wiley.
DiPrete, T.A., & Buchmann, C. (2013). The Rise of Women. The Growing Gender Gap in
Education and What it Means For American Schools. New York: Russell Sage Founda-
tion.
Durrant, G.B., D’Arrigo, J., & Steele, F. (2013). Analysing interviewer call record data by
using a multilevel time event history modelling approach. Journal of the Royal Statisti-
cal Society, 176(1), 251–269.
Dutwin, D., & Lavrakas, P. (2016). Trends in telephone outcomes, 2008–2015. Survey Prac-
tice, 9(3), 1–9.
Dykema J., Stevenson, J., Klein, L., Kim, Y., & Day, B. (2012). Effects of e-mailed versus
mailed invitations and incentives on response rates, data quality, and costs in a web
survey of university faculty. Social Sciences Computer Review, 31(3), 359–370.
Erikson, R., & Goldthorpe, J.H. (1992). The Constant Flux. A Study of Class Mobility in
Industrial Societies. Oxford: Clarendon Press.
Fine, J.P., & Gray, R.J. (1999). A Proportional Hazards Model for the subdistribution of a
competing risk. Journal of the American Statistical Association, 94(446), 496–509.
Glauser, D. (2015). Berufsausbildung oder Allgemeinbildung. Wiesbaden: VS Springer.
Göritz, A. S. (2014). Determinants of the starting rate and the completion rate in online pan-
el studies. In M. Callegaro, R. Baker, J. Bethlehem, A.S. Göritz, J.A. Krosnick, & P.J.
Becker: Gender and Survey Participation 27

Lavrakas (eds.), Online Panel Research: A Data Quality Perspective. (pp. 154–170).
New York: John Wiley & Sons.
Göritz, A.S., & Stieger, S. (2009). The impact of the field time on response, retention, and
response completeness in list-based web surveys. International Journal of Human-
Computer Studies, 67(4), 342–348.
Goyder, J., & Leiper, J.M. (1985). The decline in survey response: a social values interpreta-
tion. Sociology, 19(1), 55–71.
Goyder, J., Boyer, L., & Martinelli, G. (2008). Integrating exchange and heuristic theories of
survey nonresponse. Bulletin de Méthodologie Sociologique, 92(1), 1–14.
Green, K.E. (1996). Sociodemographic factors and mail survey response. Psychology &
Marketing, 13(2), 171–184.
Green, K.E. (2014). Reluctant respondents. Differences between early, late and nonre-
sponders to a mail survey. Journal of Experimental Education, 59(3), 268–276.
Groves, R.M. (2006). Nonresponse rates and nonresponse bias in household surveys. Public
Opinion Quarterly, 70(5), 646–675.
Groves, R.M., & Couper, M.P. (1998). Nonresponse in Household Interview Surveys. New
York: John Wiley & Sons.
Groves, R.M., & Peytcheva, E. (2008). The Impact of nonresponse rates on nonresponse
bias: a meta-analysis. Public Opinion Quarterly, 72(2), 167–189
Groves, R.M., Cialdini, R.B., & Couper, M.P. (1992). Understanding the decision to partici-
pate in a survey. Public Opinion Quarterly, 56(4), 475–495.
Groves, R.M., Dillman, D.A., Eltinge, J.L., & Little, R.J.A. (2001). Survey Nonresponse.
New York: Wiley & Sons.
Groves, R.M., Singer, E., & Corning, A. (2000). Leverage–saliency theory of survey par-
ticipation. Description and an illustration. Public Opinion Quarterly, 64(3), 299–308.
Grunow, D. (2013). Gender-based division of labour and socialisation as a matter of biogra-
phy. Zeitschrift für Soziologie der Sozialisation und Erziehung, 33(4), 384–398.
Hadjar, A. (2011). Einleitung. In A. Hadjar (Ed.), Geschlechtsspezifische Bildungsungleich-
heiten (pp. 7–20). Wiesbaden: Springer VS.
Kalter, F., Granato, N., & Kristen, C. (2007). Disentangling recent trends of the second
generation’s structural assimilation in Germany. In S. Scherer, R. Pollak, G. Otte,
& M. Gangl (eds.), From Origin to Destination. Trends and Mechanisms in Social
Stratification Research (pp. 214–245). Frankfurt am Main: Campus.
Keusch, F. (2015). Why do people participate in web surveys? Applying survey participation
theory to Internet survey data collection. Management Review Quarterly, 65(January),
183–216.
Kleiner, B., Lipps, O., & Ferrez, E. (2105). Language ability and motivation among foreign-
ers in survey responding. Journal of Survey Statistics and Methodology, 3(3), 339–360.
Kreuter, F. (2013). Improving Surveys with Paradata: Introduction. In F. Kreuter (Ed.), Im-
proving Surveys with Paradata (pp. 1–9). Hoboken, NJ: Wiley.
Kwak, N., & Radler, B. (2002). A comparison between mail and web surveys: response
pattern, respondent profile, and data quality. Journal of Official Statistics, 18(2),
257–273.
Laguilles, J. S., Williams, E.A., & Saunders, D.B. (2011). Can lottery incentives boost web
survey response rates? Findings from four experiments. Research in Higher Education,
52(5), 537–533.
28 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

Lambert, P.C. (2017). The estimation and modeling of cause-specific cumulative incidence
functions using time-dependent weights. Stata Journal, 17(1), 181–207.
Leeper, T.J. (2019). Where have all the respondents gone? Perhaps we ate them all. Public
Opinion Quarterly, 83(S1), 280–288.
Marcus B., & Schütz, A. (2005). Who are the people reluctant to participate in research?
Personality correlates of four different types of nonresponse as inferred from self- and
observer ratings. Journal of Personal, 73(4), 959–984.
Noordzij, M., Leffondré, K., von Stralen, K.J., Zocali, C., Dekker, F.W., & Jager, K.J. (2013).
When do we need competing risks methods for survival analysis in nephrology? Ne-
phrology Dialysis Transplantation, 28(11), 2670–2677.
Patrick, M.E., Singer, E., Boyd, C.J., Cranford, J.A., & McCabe, S.E. (2013). Incentives for
college student participation in web-based substance use survey. Addictive Behavior,
38(3), 1710–1714.
Porst, R., & Briel, C. v. (1995). Wären Sie vielleicht bereit, sich gegebenenfalls noch einmal
befragen zu lassen? Oder: Gründe für die Teilnahme an Panelbefragungen. (ZUMA-
Arbeitsbericht, 1995/04). Mannheim: Zentrum für Umfragen, Methoden und Analysen.
Porter, S.R., & Whitcomb, M.E. (2005). Non-response in student surveys: the role of demo-
graphics, engagement and personality. Research in Higher Education, 46(2), 127–152.
Quenzel, G., & Hurrelmann, K. (2010), Geschlecht und Schulerfolg: Ein soziales Stratifi-
kationsmuster kehrt sich um. Kölner Zeitschrift für Soziologie und Sozialpsychologie,
62(1), 61–91.
Saßenroth, D. (2013). The Impact of Personality on Participation Decisions in Surveys. A
Contribution to the Discussion on Unit Nonresponse. Springer VS, Wiesbaden.
Schoeni, R.F., Stafford, F., McGonagle, K.A., & Andreski, P. (2013). Response rates in na-
tional panel surveys. AAPSS (The ANNALS of the American Academy of Political and
Social Science), 645(1), 60–87.
Schuster, N.A., Hoogendijka, E.O., Koka, A.A.L., Twiska, J.W.R., & Heymansa, M.W.
(2020). Ignoring competing events in the analysis of survival data may lead to biased
results: a nonmathematical illustration of competing risk analysis. Journal of Clinical
Epidemiology, 122(1), 42–48.
Singer, E. (2006). Introduction: Nonresponse bias in household surveys. Public Opinion
Quarterly, 70(5), 637–645.
Singer, E. (2011). Toward a benefit–cost theory of survey participation: evidence, further
tests, and implication. Journal of Official Statistics, 27(2), 379–392.
Singer, E., van Hoewyk, J., & Maher, M.P. (2000). Experiments with incentives in telephone
surveys. Public Opinion Quarterly, 64(2), 171–188.
Sixt, M. (2013). Wohnort, Region und Bildungserfolg. Die strukturelle Dimension bei der
Erklärung von regionaler Bildungsungleichheit. In R. Becker & A. Schulze (eds.), Bil-
dungskontexte. (pp. 457–481). Wiesbaden: Springer VS.
Slauson-Blevins, K., & Johnson, K.M. (2016). Doing gender, doing survey? Women’s gate-
keeping and men’s non-participation in multi-actor reproductive surveys. Sociological
Inquiry, 86(3), 427–449.
Smith, T. (1995). Trends in non-response rates. International Journal of Public Opinion
Research, 7(2), 157–171.
Steeh, C.G. (1981). Trends in nonresponse rates, 1952–1979. Public Opinion Quarterly,
45(1), 40–57.
Becker: Gender and Survey Participation 29

Stoop, I., Billiet, J., Koch, A., & Fitzgerald, R. (2010). Improving Survey Response: Lessons
learned from the European Social Survey. New York: John Wiley & Sons.
Tourangeau, R., & Plewes, T.J. (2013). Nonresponse in Social Science Surveys. Washington,
D.C.: National Academics Press.
Wenz, A., Al Baghal, T., & Gaia, A. (2021). Language proficiency among respondents: im-
plication for data quality in a longitudinal face-to-face survey. Journal of Survey Sta-
tistics and Methodology, 9(1), 75–93.
30 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

Appendix
Table A–1: Varimax-rotated three factor structure of personality traits items

Factor 1 Factor 2 Factor 3

Items (value range: 1 = disagree – 5 = agree)
Persistence Control belief Decisiveness

Persistence
I hate to leave something unfinished. 0.5882 0.0287 –0.0029
When I have made up my mind, I manage to keep it
0.7192 –0.0581 –0.0455
up.
What I‘ve started I‘ll finish. 0.7353 –0.0640 0.0038
Even when I encounter difficulties at work, I persist
0.7373 –0.0528 –0.0649
in it.
Even with a tedious task, I don‘t give up until I‘m
0.7264 –0.0167 –0.0152
done.
Control belief
I am happy to take responsibility. 0.5238 –0.1482 0.2326
It has proven to be good for me to make decisions on
0.5408 –0.1443 0.1740
my own instead of turning to fate.
When there are problems and resistance, I usually
0.5807 –0.1498 0.1827
find ways and means to assert myself.
Success depends on luck, not performance. 0.0124 0.0878 0.8260
I feel like I have little control over what happens to
–0.0059 0.2084 0.7648
me.
Decisiveness
I am very unsure of how to decide and often fluctuate
–0.0487 0.8017 0.0809
back and forth.
I let other people confuse me in my decision. –0.0871 0.7768 0.1132
After a decision, I have great doubts as to whether I
–0.0607 0.8109 0.0931
have really made the right decision.
There are so many options that I have a hard time
–0.0161 0.7772 0.0549
deciding which one to choose.
When I have made a decision, I hold on to it. 0.4613 –0.0437 –0.0111

N Minimum Maximum

Persistence 3,680 –5.0302 2.4407

Control belief 3,680 –3.0764 3.1386
Decisiveness 3,680 –4.4313 2.5783
Becker: Gender and Survey Participation 31

Table A–2: Varimax-rotated one factor structure of gender role items

Items (value range: Factor
Mean SD Minimum Maximum
1 = disagree – 5 = agree) Gender role

Female gender role: I think it‘s important for a woman…

to be employed. 0.6173
to earn much money. 0.7766
to have a successful career. 0.7675
to have children. 0.5039
to take care of the household. 0.4947
to be responsible for childcare. 0.5105

Female gender role orientation –3.54e-09 0.9655 –3.8443 2.5000

Male gender role: I think it‘s important for a man…

to be employed. 0.7244
to earn much money. 0.8087
to have a successful career. 0.8004
to have children. 0.5809
to take care of the household. 0.4666
to be responsible for childcare. 0.5724

Male gender role orientation –2.51e-09 0.9655 –4.6210 1.9810

32 methods, data, analyses | Vol. 16(1), 2022, pp. 3-32

Table A-3: Descriptive statistics (all respondents across five waves)

N % Mean SD Minimum Maximum

Gender 13,145 51.0 0 1

Waves 13,145
Wave 4 2,654 20.2 0 1
Wave 5 2,799 21.3 0 1
Wave 6 2,712 20.6 0 1
Wave 7 2,488 18.9 0 1
Wave 8 2,492 19.0 0 1
Regional opportunity structure 13,145 0.2218 0.9804 –1.6488 3.6225
Social origin (EGP) 13,145
I 1,863 14.2 0 1
II 2,453 18.7 0 1
IIIa/b 3,173 24.1 0 1
IVa/b/c 805 6.1 0 1
V/VI 2,132 15.2 0 1
VIIa/b 704 5.4 0 1
Missing values 2,024 15.4 0 1
School type 13,145
Basic requirements 3,383 25.7 0 1
Extended requirements 5,467 41.6 0 1
Advanced requirements 2,068 15.7 0 1
Missing values 2,227 16.9 0 1
Language proficiency
(z-standardized GPA) 13,145 –0.0992 0.9089 –3.3773 1.3327
Language ability
(German vs. other languages) 13,145 85.5 0 1
Persistence 13,145 0.0182 0.9192 –5.0302 2.4407
Control belief 13,145 –0.0235 0.9449 –3.0764 3.1386
Decisiveness 13,145 0.0361 0.9264 –4.4313 2.5783
Female gender role 13,145 –3.54e-09 0.9656 –3.8443 2.5000
Male gender role 13,145 –2.51e-09 0.9656 –4.6210 1.9810