COVID-19 TRACKING: GERMANY 1
1 Psychological Factors Shaping Public Responses to COVID-19 Digital Contact
2 Tracing Technologies in Germany
1* 1 2,3 4 3 Anastasia Kozyreva , Philipp Lorenz-Spreen , Stephan Lewandowsky , Paul M. Garrett ,
1 1 1 4 Stefan M. Herzog , Thorsten Pachur , and Ralph Hertwig
1 5 Center for Adaptive Rationality, Max Planck Institute for Human Development, Germany
2 6 School of Psychological Science, University of Bristol, United Kingdom
3 7 School of Psychological Sciences, University of Western Australia
4 8 Melbourne School of Psychological Sciences, University of Melbourne, Australia
* 9 Corresponding author, E-mail: [email protected] COVID-19 TRACKING: GERMANY 2
10 Abstract
11 The COVID-19 pandemic has seen one of the first large-scale uses of digital contact tracing
12 to track a chain of infection and contain the spread of a virus. The new technology has
13 posed challenges both for governments aiming at high and effective uptake and for citizens
14 weighing its benefits (e.g., protecting others’ health) against the potential risks (e.g., loss of
15 data privacy). Our cross-sectional survey with repeated measures across four samples in
16 Germany (N = 4, 357) focused on psychological factors contributing to the public adoption
17 of digital contact tracing. We found that public acceptance of privacy-encroaching
18 measures (e.g., granting the government emergency access to people’s medical records or
19 location tracking data) decreased over the course of the pandemic. Intentions to use
20 contact tracing apps—hypothetical ones or the Corona-Warn-App launched in Germany in
21 June 2020—were high. Users and non-users of the Corona-Warn-App differed in their
22 assessment of its risks and benefits, in their knowledge of the underlying technology, and in
23 their reasons to download or not to download the app. Trust in the app’s perceived
24 security and belief in its effectiveness emerged as psychological factors playing a key role in
25 its adoption. We incorporate our findings into a behavioral framework for digital contact
26 tracing and provide policy recommendations.
27 Keywords: COVID-19 | digital contact tracing | privacy | public attitudes |
28 Corona-Warn-App COVID-19 TRACKING: GERMANY 3
29 Psychological Factors Shaping Public Responses to COVID-19 Digital Contact
30 Tracing Technologies in Germany
31 Public health interventions, economic aid, and behavioral regulations have all been enlisted
32 to curb the damage of the COVID-19 pandemic (Habersaat et al., 2020; World Health
33 Organization, 2020). Before vaccines became available, behavioral measures—restricting
34 social mixing and public gatherings, tracing the contacts of those infected, and
35 implementing a combination of physical distancing rules and hygiene measures (Robert
36 Koch Institute, 2020)—were the most promising way to contain the pandemic.
37 Technological solutions have also helped stem the spread of COVID-19 (Grantz et al., 2020;
38 Oliver et al., 2020). Indeed, with the exception of the Ebola outbreak in West Africa in
39 2014–2016 (Danquah et al., 2019), the COVID-19 pandemic has seen the first large-scale
40 use of digital contact tracing for epidemiological purposes (Kahn & Johns Hopkins Project
41 on Ethics and Governance of Digital Contact Tracing Technologies, 2020). This study
42 examines the psychological factors that have contributed to the adoption of tracking apps
43 during the COVID-19 pandemic.
44 Smartphone tracking apps use GPS, telecommunication, or Bluetooth data to create a
45 list of contacts with whom a user may have been co-located (Oliver et al., 2020). This
46 contact information is stored locally on the phone or on a centralized server. If a person
47 later tests positive for COVID-19 and shares their infection status with an app, all users in
48 their contact list can be notified instantly, allowing them to self-isolate and get tested, thus
49 ideally helping to slow the spread of the virus (Ferretti et al., 2020).
50 To date, about 50 countries have introduced COVID-19 contact tracing apps; most use
51 Bluetooth tracking technologies (O’Neill et al., 2020). The Corona-Warn-App, launched in
52 Germany in June 2020, is an open-source Bluetooth-based decentralized smartphone app
53 (https://www.coronawarn.app/en) that aims to ease the burden of the pandemic on local
54 public health authorities by complementing their offline contact tracing efforts. The app
55 employs a privacy-preserving model, collecting anonymized contact data that are stored COVID-19 TRACKING: GERMANY 4
56 locally on the user’s smartphone. Like Spain’s Radar COVID app and the United
57 Kingdom’s NHS COVID-19 app, the Corona-Warn-App is based on the Exposure
58 Notification system developed by Google and Apple.
59 The epidemiological impact of digital contact tracing apps remains low (Burdinski
60 et al., 2021) or uncertain (although the NHS app has been reported to have a positive
61 impact; Wymant et al., 2021): No country has been able to prevent more widespread
62 outbreaks without implementing harsher restrictions. Most studies agree that higher
63 participation rates are needed to increase the efficacy of digital contact tracing. In most
64 countries, the uptake of digital contact tracing apps has not reached the target of 60% (see
65 Appendix Figure A1 and Table A1), a critical threshold derived from early simulation
66 models (Hinch et al., 2020; Whitelaw et al., 2020). However, more recent simulation studies
67 suggest that even levels of adoption above 20% can have a mitigating impact (Aleta et al.,
68 2020; Bianconi et al., 2021); recent evidence suggests this to be the case in the United
69 Kingdom (Wymant et al., 2021). It is therefore crucial to understand the reasons
70 underlying people’s decision to download and use the app. Only then will it be possible to
71 develop measures that can help to make digital contact tracing an effective long-term
72 epidemiological measure (Colizza et al., 2021).
73 The effectiveness of digital contact tracing technologies depends on a combination of
74 related but distinct factors (see Figure1 for a graphical representation; see also Colizza
75 et al., 2021; Rodríguez et al., 2021), including (1) functionality: the app’s architecture
76 (e.g., which protocol or exposure notification system it uses) and its privacy and risk
77 models; (2) integration: how the app is integrated into a larger environment such as the
78 public health system and how test results are uploaded (e.g., via QR codes); (3)
79 communication: including how the risks and benefits of the app are communicated to the
80 public; (4) usage: including adoption of the technology (number of downloads), continuous
81 and correct use of the app (e.g., keeping it installed and keeping Bluetooth on), and
82 compliance (e.g., willingness and ability to upload test results); (5) detection: including key COVID-19 TRACKING: GERMANY 5
83 effectiveness metrics such as the number of positive test results uploaded as a proportion of
84 all clinically diagnosed infections in the population, the app’s overall detection rate (i.e.,
85 the proportion of an infected person’s contacts who are notified by the app—including
86 contacts unknown to the infected individual), and detection accuracy (i.e., the proportion
87 of detected contacts that is free from both false positives and false negatives; see Redmiles,
88 2020); and (6) response: complying with risk warnings and taking appropriate measures
89 (e.g., taking a test or self-isolating) following risk exposure notification.
Digital Contact Tracing Effectiveness
- App’s architecture Integration within - Public - Adoption (downloads) - positive test results - Follow-up - App’s privacy the local health communication - Adherence uploaded behavior (testing, model system, including - Marketing (continuous usage) - N of close and self-isolation) - App’s risk model testing and campaigns - Compliance hidden contacts uploading of test - Media coverage (uploading test results) traced results - Accuracy and speed
FUNCTIONALITY INTEGRATION COMMUNICATION USAGE DETECTION RESPONSE
Figure 1: Factors contributing to the effectiveness of digital contact tracing technologies. Expanded based on the analysis by Rodríguez et al. (2021).
90 Here, we focus on the adoption of digital contact tracing technologies. Numerous
91 factors can influence people’s decision to download and use an app. For instance, a study
92 in Germany revealed higher adoption rates of the Corona-Warn-App among respondents
93 with a higher risk of severe illness, respondents who follow behavioral guidelines (e.g.,
94 wearing a mask), and respondents who trust the national government, the healthcare
95 system, and science in general (Munzert et al., 2021). A study from France likewise found
96 that higher trust in government is associated with higher acceptability and increased use of
97 contact tracing apps (Guillon & Kergall, 2020); similar findings have been reported for the
98 United Kingdom (Lewandowsky et al., 2021).
99 A U.S. study on the willingness to adopt warning apps has shown, using hypothetical
100 scenarios, that people consider both the risks and the benefits of such technologies
101 (Redmiles, 2020). Benefits include knowing about one’s risk exposure, feeling altruistic, COVID-19 TRACKING: GERMANY 6
102 and protecting others; risks include privacy costs and costs of mobile data use. Another
103 U.S. study found that people value both accuracy and privacy in a tracking app (Kaptchuk
104 et al., 2020). An international study likewise highlighted the importance of privacy
105 concerns, finding that 37% of participants would not download an app even if it protected
106 their data perfectly (Simko et al., 2020). At the same time, studies in Australia (Garrett,
107 White, et al., 2021), the United Kingdom (Lewandowsky et al., 2021), and among young
108 adults in Taiwan (Garrett, Wang, et al., 2021) have shown high acceptance of potential
109 tracking technologies, especially in the presence of privacy-preserving conditions. Other
110 studies and opinion pieces have highlighted the crucial role that privacy plays in public
111 adoption of COVID-19 tracking technologies (Cho et al., 2020; Hart et al., 2020).
112 This study focuses on Germany; it draws on survey data collected by an international
113 consortium that has conducted parallel representative surveys in countries including
114 Australia (Garrett, White, et al., 2021), Taiwan (Garrett, Wang, et al., 2021), and the
115 United Kingdom (Lewandowsky et al., 2021). Our study investigated two main research
116 questions (preregistered at https://osf.io/6mkag): (1) Which factors influence the public
117 acceptance of governmental use of location tracking data in an emergency? This includes
118 the question of how people perceive location tracking technologies, including their data
119 privacy and effectiveness. (2) How did people’s attitudes differ across the pandemic?
120 Repeated measurements in cross-sectional samples allowed us to compare attitudes to
121 hypothetical scenarios in the early waves with attitudes to and adoption of Germany’s
122 Corona-Warn-App, which was launched after the second of our four waves of assessment
123 (Figure2). Our third preregistered research question, which took a crosscultural
124 perspective, will be addressed in a forthcoming international project report.
125 Our study was conducted during the first 8 months of the pandemic in Germany
126 (March to November 2020). In each of the four waves, we examined how acceptable
127 respondents found a range of privacy-encroaching measures (e.g., giving the government
128 access to medical records, tracking people’s locations using mobile phone data, and COVID-19 TRACKING: GERMANY 7
COVID−19 cases & deaths in Germany: March−December 2020
25000 Wave 1 Wave 2 Wave 3 Wave 4 Wave
20000
15000
10000
5000
0 (country−wide "contact ban") Start lockdown of first Easing of restrictions Corona−Warn−App launches Germany ("light") Start of second lockdown
08−Apr 18−Apr 28−Apr 07−Jul 17−Jul 27−Jul 05−Oct 15−Oct 25−Oct 09−Mar 19−Mar 29−Mar 08−May 18−May 28−May 07−Jun 17−Jun 27−Jun Date06−Aug 16−Aug 26−Aug 05−Sep 15−Sep 25−Sep 04−Nov 14−Nov 24−Nov 04−Dec 14−Dec Figure 2: Rolling 7-day averages of daily reported COVID-19 cases (gray) and deaths (red) in Germany between March and December 2020. Data were collected at four points of measurement: Wave 1 (30–31 March), Wave 2 (17–22 April), Wave 3 (25 August–03 September), and Wave 4 (02–08 November). The timeline includes key policy decisions and the launch of the Corona-Warn-App.
129 temporarily relaxing data protection regulations). In the first two waves, respondents were
130 presented with one of three hypothetical scenarios representing different degrees of privacy
131 invasion. Each scenario described a hypothetical tracking app and the related policies (e.g.,
132 the government is required to delete all data collected by the app after 6 months). The last
133 two waves probed respondents’ attitudes toward the Corona-Warn-App itself (see
134 Appendix Table B4 for descriptions of the scenarios). We also collected a variety of
135 attitude measures, such as worldviews and COVID-19 risk perceptions, in order to identify
136 potential psychological predictors of policy acceptance (see Methods for details). Strengths
137 of our approach include the ability to compare attitudes toward three hypothetical
138 scenarios (in the earlier waves) with actual adoption rates of the Corona-Warn-App (in the
139 later waves). Our cross-sectional study with large representative online samples thus allows
140 us to identify the psychological factors behind the adoption of digital contact tracing
141 technologies, and to examine how those factors change over time.
142 We discuss our results in the light of research on the perceived risks and benefits of new
143 technologies and incorporate these insights into a behavioral framework for digital contact COVID-19 TRACKING: GERMANY 8
144 tracing (based on the behavior change wheel by Michie et al., 2011). Our adaptation of
145 this framework highlights how the three components of capability, opportunity, and
146 motivation contribute to the adoption of COVID-19 tracking technologies. We conclude by
147 offering policy recommendations aiming to encourage the public to adopt
148 privacy-preserving contact tracing apps.
149 Results
150 In order to address our two main research questions of which factors influence the
151 public response to digital contact tracing and how attitudes differ across the pandemic, we
152 ran analyses addressing the following questions: (1) How have people’s risk perceptions of
153 COVID-19 changed over the course of the pandemic? (2) How have people’s attitudes
154 towards various privacy-encroaching measures changed over the course of the pandemic?
155 (3) How acceptable do people find various types of tracking technologies? And how do
156 these attitudes compare to download rates for the Corona-Warn-App? (4) How do people
157 rate various measures of the effectiveness and risks of these technologies? (5) What are the
158 most important reasons for people to download or not download the Corona-Warn-App?
159 (6) Which factors are most predictive of app adoption and intention to download?
160 COVID-19 Risk Perceptions
161 As Figure3 shows, the clear majority of participants indicated that they thought the
162 virus posed a moderate to severe threat to the general population: The number of
163 participants stating that the virus’s severity for the population was “somewhat,” “very,” or
164 “extremely” high ranged from 84% to 96% across the four waves, with severity ratings
165 increasing along with infection rates. Median values reached the “very severe” level in
166 three of the four study waves. For all other risk perception items, median values were at
167 the “somewhat severe” level throughout the four study waves. COVID-19 TRACKING: GERMANY 9
COVID−19 risk perception Infection rates Covid−19 cases in Germany: Severity for German population Harm to your health if infected Concern you will be infected Concern someone you know will be infected March−December 2020 55
40 33 30 31 26 24 25 22 22 21 20 21 Wave 1 (30−31.Mar) 20 1 Wave 19 13 11 9 10 3 4 Wave 2 (17−22.Apr) 1 0
44 40 39 38
30 30 27 27 25 22 20 2 Wave 18 17 15 13 12 12 9 8 6 6 2 0
% of respondents 42 40 35 29 29 30 29 29 26 24 20 20 3 Wave 17 17 Wave 3 (25.Aug−03.Sept) 14 12 12 9 7 8 7 4 0
51
40 35 31 30 25 26 26 23 24 20 20 4 Wave 19 Wave 4 (02−08.Nov) 15 15 14 11 10 8 8 6 3 0 Very Very Very Very Slightly Slightly Slightly Slightly Not at all Not at all Not at all Not at all Extremely Extremely Extremely Extremely Somewhat Somewhat Somewhat Somewhat Figure 3: Perceived risks from COVID-19 across the four waves of the study. Barplots show the distribution of responses to the four items (with percentages). Boxplots show the interquartile range (IQR; responses between the 25th and 75th percentiles); the black line within each box indicates the median value. Lower and upper whiskers extend from the hinge to the smallest and largest values within 1.5 × IQR. Plot on the right shows COVID-19 infection reported in Germany over the study waves (rolling 7-day averages of daily reported COVID-19 cases). For wordings of the risk items, see Appendix Table B2.
168 The proportion of people who believed that the virus threatened their health
169 “somewhat” remained stable (between 27% and 31%), while the proportion of people who
170 thought the threat was “very” or “extremely” high fluctuated, tending toward higher
171 numbers with time (March: 35%, April: 30%, September: 46%, November: 41%). Overall,
172 on the aggregate level, people were more concerned about the health of others than about
173 their own health as indicated by the distribution of responses in the barplots (Figure3).
174 Across all four waves, on the individual level (within respondents), the majority of
175 participants were equally concerned about the risk of infection to themselves and to others
176 (Appendix Figure A2). However, over time, people showed increased concern for
177 themselves, reflected in the rising proportion of people concerned equally for themselves COVID-19 TRACKING: GERMANY 10
178 and others (March and April: 49%, September and November: 59%) and the decreasing
179 proportion of people reporting more concern for others (March and April: 43–44%,
180 September and November: 33%). The proportions of respondents who indicated more
181 concern for themselves than for others remained stable (7–8%) across all four waves.
182 Acceptability of Privacy-Encroaching Measures
183 As Figure4 shows, the acceptability of privacy-encroaching measures that could
184 hypothetically be implemented by the government (e.g., temporarily suspending data
185 protection regulations) was fairly high, but tended to decrease over time.
186 Whereas respondents’ risk perceptions tracked the pandemic’s development in
187 Germany—that is, perceived risk for the country as a whole was higher at the end of
188 March and April (Waves 1 and 2, respectively) and November (Wave 4), when infections
189 were rising (Figure3)—respondents’ attitudes toward privacy-encroaching measures
190 followed a different pattern. After the initial shock of the pandemic, the acceptability of all
191 measures tended to decrease from thereon, as reflected in the decreasing percentages of
192 people who deemed these measures “very” or “somewhat” acceptable. Cochran–Armitage
193 tests for trend corroborate the decreasing rates of acceptability across four study waves,
194 especially for the first four measures in Figure4( p < 0.001). For the other two measures
195 (temporarily relaxing data protection regulations and collecting data on contacts and
196 interactions) there was an indication of a decreasing trend (p = 0.01 and p = 0.02,
197 respectively; see Appendix Table A2 for detailed results).
198 Closer inspection revealed that, within this overall trend of decreasing acceptability
199 over time, there were two distinct patterns of attitudinal change: a steep gradual decrease
200 in acceptability and a pattern that more closely mirrored the development of the pandemic.
201 Measures such as allowing access to medical records or location tracking data fall into the
202 first pattern. Granting the government access to citizens’ medical records was deemed
203 “very” or “somewhat” acceptable by 68% of participants in Wave 1; this number dropped COVID-19 TRACKING: GERMANY 11
General acceptability of six privacy−encroaching measures
100
P−value for trend <0.001 <0.001 <0.001 <0.001 0.01 0.02 75
69 68 65 60 59 60 59 50 54 55 52 52 50 49 48 47 48 48 45 46 45 45 41 37 35 25 Acceptability (% of respondents)
0
Data on infections & Access to Location Notification when people Relaxation of data Data on contacts immunity status medical records tracking data leave quarantine protection regulations & interactions
Wave 1 Wave 2 Wave 3 Wave 4
Infection rates Covid−19 cases in Germany: March−December 2020 Wave 2 (17−22.Apr) Wave Wave 1 (30−31.Mar) Wave Wave 4 (02−08.Nov) Wave Wave 3 (25.Aug−03.Sept) Wave
Figure 4: Acceptability of privacy-encroaching measures in Germany across the four waves of the study. Acceptability scores represent the total percentage of participants who chose the response options “very acceptable” or “somewhat acceptable” to the question “How acceptable is it for the government to take the following measures to limit the spread of the virus during the COVID-19 pandemic?” Error bars are 95% confidence intervals computed with the R function prop.test; results of this test are reported in Appendix Table A3. p values from Cochran–Armitage tests for trend shown in gray boxes above the barplots; see Appendix Table A2 for detailed results. Plot at the bottom shows COVID-19 infection reported in Germany over the study waves (rolling 7-day averages of daily reported COVID-19 cases). See Appendix Table B9 for the wording of the items.
204 in each wave, reaching just 35% in Wave 4 despite the rise in infections and new lockdown
205 measures at that time. Acceptability of collecting people’s location tracking data followed
206 the same pattern (Figure4). In contrast, measures such as collecting data on people’s
207 infections and immunity status or their contacts and interactions seemed to be more
208 responsive to the pandemic’s development and associated risk perceptions (see also
209 Appendix Figure A3 for correlations between acceptability of privacy-encroaching measures COVID-19 TRACKING: GERMANY 12
210 and COVID-19 risk perceptions). For example, 49% of respondents found collecting data
211 on people’s contacts and interactions to be “somewhat” or “very” acceptable at the end of
212 March, during the first phase of the pandemic in Germany. This decreased over the next
213 two waves, then rose to 45% in November, mirroring the increase in infections in Germany
214 at that time. Note, however, that acceptance of this measure stayed below 50% across all
215 four waves of the study.
216 These two distinct patterns suggest that people may have adapted their attitudes
217 during the pandemic, distinguishing between measures they deem more appropriate (e.g.,
218 collecting data on infections and immunity status) and measures they initially deemed
219 acceptable but whose acceptability gradually decreased over time (e.g., granting access to
220 medical records or location tracking data).
221 Acceptability of Tracking Technologies
222 We found relatively high levels of acceptance of the three tracking technologies
223 presented in the hypothetical scenarios in Waves 1 and 2 (mild, severe, and Bluetooth).
224 Acceptability of all three was above 50% in both waves. There were no large differences
225 between proportions of respondents who found acceptable the three hypothetical scenarios
226 (Figure5).
227 Surprisingly, although the severe scenario was deemed least acceptable (56% in Wave 1
228 and 55% in Wave 2 compared to 61% and 64%, respectively, for the mild scenario and 59%
229 in Wave 2 for Bluetooth), its acceptance level was not particularly low. The differences
230 between scenarios virtually disappeared when respondents considered follow-up options
231 (e.g., deletion of all data after 6 months or opting out of data collection). The reported
232 percentage of downloads of the Corona-Warn-App in our samples was smaller (36% and
233 41% in Waves 3 and 4, respectively) than the acceptability of the hypothetical scenarios.
234 This low number of reported downloads is consistent with the actual download rates for
235 the Corona-Warn-App in Germany (currently estimated at about 30% of the population; COVID-19 TRACKING: GERMANY 13
Acceptability of Tracking in Three Scenarios and Corona−Warn−App Downloads
Severe Mild Bluetooth Corona−Warn−App 100
75 79 80 79 74 71 72 69 69 67 50 64 61 59 56 55 55 52 41 % of respondents 25 36
0
Acceptability With follow up: With follow up: Acceptability With follow up: With follow up: Acceptability With follow up: With follow up: Downloads of the scenario delete data opt out of the scenario delete data local data of the scenario delete data future downloads
Wave 1 Wave 2 Wave 3 Wave 4 Figure 5: Acceptability of hypothetical tracking technologies and numbers of Corona-Warn-App downloads. For the hypothetical scenarios, the first column displays baseline acceptability ratings; the other two columns display acceptability under varying conditions: with the requirement that all data be deleted and tracking stopped after 6 months; with an “opt out” option (severe scenario); and with data being stored locally on the user’s phone (mild scenario). For Corona-Warn-App usage, the data show current downloads and intentions to download. Error bars are 95% confidence intervals computed by R function prop.test. Full results are reported in Appendix Table A4. For items, see Appendix Table B5; for descriptions of scenarios, see Appendix Table B4.
236 see Figure A1). The somewhat higher download rate reported here might be explained by
237 the demographics of our sample, which was skewed toward online users aged 18 years or
238 older. When respondents in Waves 3 and 4 were asked whether the Corona-Warn-App
239 should be mandatory, only 29% and 30%, respectively, said yes (Appendix Table A5). This
240 could indicate that people were less likely to find tracking technologies acceptable at later
241 stages of the pandemic (consistent with the trend in Figure4); it could also indicate a
242 difference in participants’ perceptions of hypothetical scenarios and the actual app: The
243 latter leaves less room for interpretation when participants answer the survey questions.
244 Perceptions of Effectiveness and Risks of Tracking Technologies
245 Figure6 displays perceptions of the effectiveness and risks of the tracking technologies
246 and policies presented. COVID-19 TRACKING: GERMANY 14
Perception of risk and effectiveness in each scenario and in case of the Corona−Warn−App
Severe
Very much
Quite a lot
Moderately
A little
Hardly
Not at all
Reduce Ease of Only Trust: Data Reduce Return to Risk How Sensitive Trust: Privacy Trust: Security User Control Likelihood Declining Necessary for Pandemic Spread Activity of Harm Is Data Protection from 3rd Party Over Data to Contract Participation Data Collection Only
Mild
Very much
Quite a lot
Moderately
A little
Hardly
Not at all
Reduce Ease of Only Trust: Data Others' Ability Reduce Return to Risk How Sensitive Trust: Privacy Trust: Security User Control Likelihood Declining Necessary for Pandemic to Download Spread Activity of Harm Is Data Protection from 3rd Party Over Data to Contract Participation Data Collection Only the App
Bluetooth Level of Confidence Level Very much
Quite a lot
Moderately
A little
Hardly
Not at all
Reduce Ease of Only Trust: Data Others' Ability Reduce Return to Risk How Sensitive Trust: Privacy Trust: Security User Control Likelihood Declining Necessary for Pandemic to Download Spread Activity of Harm Is Data Protection from 3rd Party Over Data to Contract Participation Data Collection Only the App
Corona−Warn−App
Very much
Quite a lot
Moderately
A little
Hardly
Not at all
Reduce Reduce Reduce Return to Return to Ease of Only Trust: Data Others' Ability Risk How Sensitive Trust: Privacy Trust: Security User Control Likelihood Spread Spread Activity Activity Declining Necessary for Pandemic to Download of Harm Is Data Protection from 3rd Party Over Data to Contract Past Future Past Future Participation Data Collection Only the App
Wave 1 Wave 2 Wave 3 Wave 4 Figure 6: Perception of effectiveness and risks of the tracking policy in each scenario. Boxes show the interquartile range (IQR; responses between the 25th and 75th percentiles); the black horizontal line inside the boxes indicates the median value. Lower and upper whiskers extend from the hinge to the smallest and largest values within 1.5 × IQR. Individual responses are jittered horizontally and vertically. For items, see Appendix Table B6; for descriptions of scenarios, see Appendix Table B4.
247 Participants understood that the severe scenario posed a greater risk to data privacy
248 and data sensitivity, control over user data, and ability to decline participation than the
249 other hypothetical scenarios. At the same time, they judged the potential effectiveness of
250 the severe scenario to be on the same level as in the other two scenarios (mild and COVID-19 TRACKING: GERMANY 15
251 Bluetooth). It is therefore puzzling that the acceptability of the severe scenario was almost
252 on par with the other two (see Figure5), even though participants thought the risk to
253 privacy protection and the level of intrusion in citizens’ lives was much higher.
254 Figure6 also shows that although participants thought the Corona-Warn-App
255 presented only a low risk of harm, they were pessimistic about its effectiveness, including
256 its ability to reduce the spread of the virus and to help people return to their normal
257 activities. This pessimism toward the Corona-Warn-App was stronger than that toward
258 the hypothetical scenarios presented in earlier waves of the study. Moreover, participants
259 showed only moderate levels of trust in the Corona-Warn-App’s security (including trust in
260 collection of only necessary data and only for the purposes related to the pandemic, trust
261 in privacy protection and trust in data security from third parties)—closest to that found
262 in the mild scenario, but higher than that in the severe and Bluetooth scenarios. Given
263 that the technology in the Bluetooth scenario was attributed to Apple and Google, while
264 the Corona-Warn-App was released by the German government, the lower level of trust in
265 the Bluetooth scenario may be due to a lack of trust in commercial corporations and their
266 standards of data protection.
267 People’s perceptions of the risks and benefits of the Corona-Warn-App differed
268 depending on whether or not they already had downloaded the app. On the aggregate
269 level, app users judged its risk of harm to be very low and trusted the app’s security a lot
270 (see top part of Figure7.A). In contrast, non-users’ aggregate ratings of both risk of harm
271 and trust in security were both at the same level (median = “a little”). This pattern was
272 also present on the individual level (within respondents), where, for instance, the majority
273 of users (86% and 82% in Waves 3 and 4, respectively) rated the app’s security higher than
274 its risk of harm. At the same time, only about half (50% and 51%) of non-users rated the
275 app’s security higher than its risk of harm, and even fewer non-users (38% and 39%) rated
276 its effectiveness higher than its risk of harm. The opposite was true for app users, among
277 whom the majority (77% and 70%) rated the app’s effectiveness higher than its risk of COVID-19 TRACKING: GERMANY 16
278 harm (see Appendix Figure A4).
A. Effectiveness and risks of the Corona−Warn−App (CWA) across respondents
CWA downloaded: Yes CWA downloaded: No
Very much
Quite a lot
Moderately
A little
Hardly
Not at all
mean aggregate value per respondent value mean aggregate Trust Belief Risk Trust Belief Risk in security in effectiveness of harm in security in effectiveness of harm
B. What technology does the Corona−Warn−App (CWA) use?
CWA downloaded: Yes CWA downloaded: No 100
76 75 65
50 35 36 32 35 26 26 25 18 22
% of respondents 9 3 4 3 4 7 0
Mobile Location Mobile Location Do not know Bluetooth Do not know Bluetooth tower data data tower data data
Wave 3 Wave 4
Figure 7: Users and non-users of the Corona-Warn-App: Perception of effectiveness, risks, and understanding of technology. Panel A: Perception of effectiveness and risks of the Corona-Warn-App for two groups: those who had downloaded the app and those who had not. The variables “Trust in security” and “Belief in effectiveness” represent combined and averaged measures from variables presented in Figure 6. The variable “Trust in security” combines four variables: “Only Necessary Data Collection,” “Trust: Privacy Protection,” “Trust: Security from 3rd Party.” The variable “Belief in effectiveness” combines three variables: “Reduce Likelihood to Contract,” “Reduce Spread Future,” “Return to Activity Future.” “Risk of harm” is a single variable. For items, see Appendix Tables B6 and for all aggregate measures see B10. Boxplots show distributions of aggregate (mean) values. Boxes show the interquartile range (IQR; responses between the 25th and 75th percentiles); the black horizontal line inside the boxes indicates the median value. Lower and upper whiskers extend from the hinge to the smallest and largest values within 1.5 × IQR. Individual responses are jittered horizontally and vertically. Panel B. Understanding of the Corona-Warn-App technology. Participants who had downloaded the app were much more likely to give the correct answer: Bluetooth. COVID-19 TRACKING: GERMANY 17
279 Knowledge of the technology used by the Corona-Warn-App also differed depending on
280 whether or not respondents had downloaded it. As Figure7.B shows, many non-users did
281 not understand how the Corona-Warn-App works: only 35% (Wave 3) and 26% (Wave 4)
282 of them knew that it uses Bluetooth technology—relative to 76% (Wave 3) and 65% (Wave
283 4) of respondents who had downloaded the app.
284 To delve deeper into these differences and their impact on people’s adoption of digital
285 contact tracing technologies, in our follow-up questions and analyses we explored potential
286 drivers of the decision to download or not download the Corona-Warn-App.
287 Corona-Warn-App: Reasons for and Against Download
288 The relatively low uptake of the Corona-Warn-App could be due to a variety of factors.
289 To explore these factors, we presented respondents with several possible reasons to
290 download or not download the Corona-Warn-App (multiple selections allowed; Figure8).
Reasons to download Corona−Warn−App Reasons not to download Corona−Warn−App
19 34 Belief it is not effective 22 To protect my health 15 35 Concerns: Privacy 18
31 12 Concerns: 3rd party access To protect the health of others 14
33 12 Lack of trust in government 13
18 10 To follow government recommendations Concerns: Normalizing government tracking 10 17 10 Concerns: Civil liberties 9 9 5 To return to normal activities Don't own a smartphone 8 6 4 Phone too old 6 6
To help the economy Other reasons 5 7
Belief virus is not dangerous 5
Other reasons 3 4 Concerns: Battery usage 4
0 10 20 30 40 50 0 10 20 30 40 50 % of respondents % of respondents
Wave 3 Wave 4 Wave 3 Wave 4 Figure 8: Self-reported reasons to download or not download the Corona-Warn-App. Panels show results of multiple-choice items in Waves 3 and 4. By design, there are more response options for both questions in Wave 4 than in Wave 3. See Appendix Table B7 for the wording of the items. COVID-19 TRACKING: GERMANY 18
291 The results indicate that people’s main reason for downloading the app was their desire
292 to protect their health and the health of others. The two leading reasons for people not
293 downloading the app were the belief that the app is not effective and privacy concerns.
294 Concerns about third-party access and lack of trust in the government also played a role.
295 To analyze responses to the open-ended question about reasons to download or not
296 download the app, we extracted unigrams (i.e., individual words) from the responses and
297 counted their overall frequencies as well as their co-occurrences within responses. Figure9
298 shows the resulting co-occurrence networks for reasons to download the app from 477
299 individual responses (panel a) and reasons to not download the app from 530 individual
300 responses (panel b). Clusters of frequent words indicate the main arguments.
join
phone
good
health department
prosecution
case
reasonschain of infection for: reasons against: a) government b)
support hoped
suitable
iphone contacts none
make sense spreading useful
effort
recommendation
easier
old helps or for this
trap
control
warn let
download combat own
leave old important do possible contacts find hardly
benefit
many
necessary follow-up hold infection use
apartment smartphone
as possible personal time
at
feel
use possibility contact work leave
thought
near cell phone home think helpful good people help warned bring contain operating system
come maybe rarely feel chains of infection personal goes
tested
someone
need ahn timely location device help voluntarily at home opinion recognize concern good
afford
would be pandemic thereby bluetooth useful virus moreover
contribution protection infected even problems smartphone containment dissemination therefore 24
try inform memory never
really safer 100 to help constantly
correct
bring people infected better family people correct
times storage space
contribute
virus simply
warn goes contribute eh
estimate mature out life works responsibility who at all rules infection proximity if necessary hold download pointless useful test little
support protect convinced if person enough battery use someone makes power 19
friends population brings possible belief find already uses unnecessary part always covid
used endangered safe contact thereby informed download
more prevent people need tested gives would like apps safe
completely
control fear whole more infected useless
would like important bullshit risk group
handle convinced infected knows from it on the road belong sense know quarantine white crazy
fear corona infected
because of activates positive quite trust corona persons
government see
risk was monitoring few safety as long as
infection environment
data many HEALTH OFFICE
work positive possibly encounter unsafe get warned
own was to infect
come data protection
someone
perfect helps
keep spied give
always
go
recommended
tested let to it
gives therein little people think trust about
enter interest feel
simply
wrong
infected
encounters look health
contained
results
see
additional trust
coronavirus
unreliable
user
also
works
help
Figure 9: Self-reported reasonsresult to download or not download the Corona-Warn-App in an open-response question (Wave 4 only; reasons for: N = 477; reasons against: N = 530). Co-occurrence networks of unigrams from reasons foralready (panel a) and against (panel b) downloading the app. Connections appear whenever two words were used by the same participant. Node and font sizes and color code are proportional to the absolute frequency of the corresponding word; nodes are positioned using a spring layout. Only unigrams that appeared at least three times in the responses are shown. Translation on a unigram basis via DeepL.com; visualization via Gephi (Bastian et al., 2009).
301 Reasons for downloading the app include protecting oneself and others (around the
302 term “protect”), being informed about infections in one’s environment (“informed” and
303 “contact”), and helping to mitigate the pandemic (“pandemic,” “contain,” and “help”).
304 Reasons against downloading the app include technical issues (“smartphone”), data privacy COVID-19 TRACKING: GERMANY 19
305 (“data protection” and “trust”), problems of functionality (“works”), and doubts around
306 how useful and necessary the app is (“hold” and “brings [nothing]”, translated from
307 German where these words are part of expressions that mean “I consider” and “pointless”).
308 Overall, the reasons for not downloading the app are slightly more diverse than the reasons
309 for downloading it; this was also the case for the multiple-choice question (Figure8). The
310 main difference between the multiple-choice and the open-ended responses is the more
311 prominent role of problems with smartphones (reasons against) and of being informed
312 (reasons for) in the open responses.
313 Corona-Warn-App: Predictors of Download
314 To further examine why people chose to download the Corona-Warn-App, we fitted a
315 logistic regression model using a set of independent variables measured in the survey as
316 predictors for the dependent variable of downloading the Corona-Warn-App (Figure 10).
317 Once again, trust in the app’s security and perceived effectiveness emerged as leading
318 positive predictors of downloading the app. These two variables represent combined
319 measures made up of variables presented in Figure6. The variable “Trust in CWA
320 security” included items asking respondents how much they trusted the app to ensure
321 individuals’ privacy (“Trust: Privacy Protection”) and to only collect and use the
322 Corona-Warn-App data to deal with the pandemic (“Only Necessary Data Collection” and
323 “Trust: Data for Pandemic Only”), as well as how secure they thought the data collected
324 by the app actually is (“Trust: Security from 3rd Party’)’ . The variable “Perception of
325 CWA effectiveness” included items asking for people’s assessment of whether the app will
326 help reduce the virus’s spread (“Reduce Spread Future”), reduce their likelihood of coming
327 into contact with the virus (“Reduce Likelihood to Contract”), and help them return to
328 their normal activities (“Return to Activity Future”). This variable therefore represents
329 people’s assessment of the app’s potential to impact the course of the pandemic and help
330 them personally (see Appendix Table B10 for all predictor variables). COVID-19 TRACKING: GERMANY 20
Predictors of Corona−Warn−App (CWA) download
Trust in CWA security
Perception of CWA effectiveness
Acceptance of privacy limits
COVID−19 risk perception
Belief in conspiracies
Social media use & trust
Government announcements use & trust
Wave Approval of German pandemic response
Wave 3 Find government guidelines helpful Wave 4
Attitudes to science and technology
Libertarianism
CWA risk of harm perception
Age
Gender: male (vs. female)
Education: low (vs. medium)
Education: high (vs. medium)
−1 0 1 2 3 Estimate Figure 10: Logistic regression models predicting Corona-Warn-App download in Waves 3 and 4. Dependent variable: downloaded the app (yes/no). Coefficients: measures from the survey (e.g., combined score for trust in CWA security; combined score for conspiracy beliefs; see Appendix Table B10). Horizontal point ranges show point estimates and 95% confidence intervals for each predictor. Education was dummy coded with the reference level “medium,” yielding two coefficients: low (vs. medium) and high (vs. medium) education. Following Gelman (2008), we standardized all continuous variables by two standard deviations (SD) and mean centered the binary gender variable. This way a 2-SD change in a continuous predictor variable is approximately equivalent to a change of category in a roughly balanced binary predictor variable (e.g., gender). In a logistic regression model, a slope reflects the relative change in log odds (while keeping all other predictors at their average values). Appendix Table A6 summarizes the regression results for these two models. Appendix Figures A6 and A7 display Pearson correlations for all variables in the regression model. Appendix Figure A8 provides an alternative arrangement of the same results, where the respective model for each wave is shown in a separate panel.
331 As trust in the app’s security and perceived effectiveness emerged as strong predictors
332 of having downloaded the Corona-Warn-App, we used the same modeling approach to
333 assess predictors of both of these variables separately. Appendix Figures A10 and A11
334 show the results of the respective linear regression models. Acceptance of privacy limits
335 during the pandemic (a combined measure summarizing all six items discussed under
336 “Acceptability of Privacy-Encroaching Measures” and in Figure4) and trust in science and COVID-19 TRACKING: GERMANY 21
337 government guidelines emerged as moderate positive predictors of both variables.
338 Furthermore, believing in conspiracy narratives and perceiving the Corona-Warn-App as
339 harmful were associated with lower trust in the app’s security, but not with the perception
340 of its effectiveness. Demographic factors such as a high level of education and identifying
341 as male—but not age—also emerged as positive predictors of having downloaded the
342 Corona-Warn-App. As Appendix Figure A5 shows, proportions of respondents who
343 reported having downloaded the app were higher at high and medium levels of education
344 than at a low level: For instance, 46% of participants with a university degree had
345 downloaded the app, relative to just 29% (Wave 3) and 35% (Wave 4) of participants with
346 a low level of education. Slightly more male respondents (Wave 3: 44%, Wave 4: 40%)
347 than female respondents (Wave 3: 38%, Wave 4: 32%) reported having downloaded the
348 Corona-Warn-App.
349 Finally, we used logistic regression models to analyze the intention to download the
350 Corona-Warn-App among respondents who reported that they had not already done so.
351 Perceived effectiveness of the app emerged as the strongest predictor of intention to
352 download (Figure 11) in both Waves 3 and 4. This finding is in line with self-reported
353 reasons for not downloading the app, where effectiveness concerns (along with privacy
354 concerns) played a large role (see Figure8). COVID-19 TRACKING: GERMANY 22
Predictors of intention to download Corona−Warn−App
Trust in CWA security
Perception of CWA effectiveness
Acceptance of privacy limits
COVID−19 risk perception
Belief in conspiracies
Social media use & trust
Government announcements use & trust
Wave Approval of German pandemic response
Wave 3 Find government guidelines helpful Wave 4
Attitudes to science and technology
Libertarianism
CWA risk of harm perception
Age
Gender: male (vs. female)
Education: low (vs. medium)
Education: high (vs. medium)
−1 0 1 2 3 Estimate Figure 11: Logistic regression models for intention to download the Corona-Warn-App (Waves 3 and 4). Dependent variable: intention to download the app in the future (yes/no). Coefficients: measures from the survey (e.g., combined score for trust in app security; combined score for conspiracy beliefs; see Appendix Table B10). Horizontal point ranges indicate point estimates and 95% confidence intervals for each predictor. Education was dummy coded with the reference level “medium,” yielding two coefficients: low (vs. medium) and high (vs. medium) education. Following Gelman (2008), we standardized all continuous variables by two standard deviations (SD) and mean centered the binary gender variable. This way a 2-SD change in a continuous predictor variable is approximately equivalent to a change of category in a roughly balanced binary predictor variable (e.g., gender). In a logistic regression model, a slope reflects the relative change in log odds (while keeping all other predictors at their average values). Appendix Table A7 summarizes the regression results for both models.Appendix Figure A9 provides an alternative arrangement of the same results, where the respective model for each wave is shown in a separate panel.
355 Discussion
356 Which Psychological Factors Contribute to Public Adoption of Digital Contact
357 Tracing Technologies?
358 Digital contact tracing apps have been introduced in many countries to help to contain
359 COVID-19 outbreaks. The apps were launched with high hopes, but have also posed a
360 number of challenges both for governments aiming at high and effective uptake and for COVID-19 TRACKING: GERMANY 23
361 citizens weighing the benefits (e.g., public and individual health) against the potential risks
362 (e.g., loss of data privacy) of these unprecedented measures. Although their
363 epidemiological impact is presently limited or uncertain (Burdinski et al., 2021; but see
364 Wymant et al., 2021), digital contact tracing apps are poised to become an established
365 epidemiological tool. Ideally, along with other measures, they can help to prevent large
366 outbreaks by relieving the burden on public health agencies. With this potential in mind,
367 it is crucial to understand which factors contribute to the public uptake of digital contact
368 tracing apps. In the present survey in Germany, we focused on psychological factors that
369 foster or inhibit the public adoption of digital contact tracing technologies.
370 Research on risk perception has identified several psychological drivers that figure in
371 people’s risk–benefit calculations and contribute to the acceptance of novel technologies
372 (see Fischhoff et al., 1978; Frey, in press; Siegrist, 2000; Slovic, 1987). A first factor is the
373 subjective perception of risks, including a risk’s severity, defined in terms of perceived lack
374 of control and catastrophic potential (“dread risks”), and novelty, defined in terms of
375 unobservable and unpredictable consequences and their potential harm (“unknown risks”;
376 see Slovic, 1987). A second factor is the perception of benefits, both personal (e.g.,
377 returning to normal activities, avoiding infection) and public (curbing the virus’s spread;
378 see also Fischhoff et al., 1978). A third and related factor is individuals’ degree of knowledge
379 and experience of a technology, which facilitates or precludes accurate assessment of its
380 risks and benefits (e.g., Gstraunthaler and Day, 2008, see also Siegrist, 2021). A fourth
381 factor is trust in the authorities that develop, deploy, or supervise the technology. The
382 trust factor becomes particularly important in the absence of knowledge about technology,
383 as a way to cope with the complexity of the task by relying on experts or relevant
384 authorities for risk management (Siegrist, 2000, 2021). An additional factor to consider in
385 risk–benefit assessments of digital contact tracing technologies is the context (e.g., the risk
386 posed by the pandemic itself) and its development (e.g., whether infection rates are
387 decreasing or increasing). For example, people might be willing to overlook risks to their COVID-19 TRACKING: GERMANY 24
388 privacy temporarily in order to mitigate the health risks posed by the pandemic. Results of
389 our study speak to several of these factors, as summarized in the following sections.
390 Privacy in the context of COVID-19 : We found that public acceptance of potential
391 privacy-encroaching measures in the context of the pandemic decreased over time. This
392 trend was especially pronounced for such hypothetical measures as granting the
393 government emergency access to people’s location tracking data and medical records (see
394 Figure4). Acceptability of all six hypothetical privacy-encroaching measures across the
395 four waves was correlated with people’s COVID-19 risk perceptions (see Appendix Figure
396 A3). Moreover, our results suggest that people might distinguish between measures they
397 considered to be more appropriate during the pandemic (e.g., collecting data on infections
398 and immunity status) and measures that were initially deemed acceptable but whose
399 acceptability gradually decreased over time (e.g., granting access to medical records or
400 location tracking data).
401 Severity and acceptability of digital tracking: Acceptability ratings for all three
402 hypothetical scenarios in Waves 1 and 2 were high, as were intentions to download the
403 Corona-Warn-App in Waves 3 and 4. Surprisingly, the details of the tracking technologies
404 presented in the scenarios made little difference to public acceptance: The severe scenario
405 involved harsh, nearly oppressive measures, while the mild and Bluetooth scenarios were
406 compatible with privacy protection standards. However, privacy measures such as having
407 all data deleted after 6 months or having the ability to opt out increased levels of
408 acceptance. Similarly high acceptance rates for all three scenarios have been observed in
409 Australia (Garrett, White, et al., 2021), the United Kingdom (Lewandowsky et al., 2021),
410 and Taiwan (Garrett, Wang, et al., 2021).
411 Taken together, these findings indicate that although people might accept certain
412 limitations to their privacy in a crisis, they are also wary of privacy-encroaching measures
413 and weigh the potential benefits of disclosing sensitive data against the potential risks (see
414 also Dienlin and Metzger, 2016). Long-term tracking solutions should therefore not rely on COVID-19 TRACKING: GERMANY 25
415 privacy-encroaching measures. Instead, they should build on privacy-preserving
416 technologies, which are more likely to be accepted in the long term.
417 Perceived benefits and risks, and people’s knowledge about the Corona-Warn-App:
418 Perceived benefits, both personal and societal, played a key role in people’s self-reported
419 reasons to download the Corona-Warn-App (Figure8). Although the perceived risk of
420 harm was low overall (Figure6), concerns about data privacy and government tracking
421 were high among non-users of the app and were the leading reported reasons not to
422 download the app(Figure8). Users and non-users of the app also showed different patterns
423 in their assessment of the app’s risks and benefits: Users judged the risk of harm to be very
424 low and reported high trust in its security; non-users, in contrast, rated both risk of harm
425 and trust in security at equally low levels (Figure7.A). Although users were, on average,
426 more optimistic about the app’s effectiveness than non-users, beliefs in the app’s
427 effectiveness were in general not high—as also observed by Munzert et al. (2021). Users
428 and non-users also differed in their knowledge of the Corona-Warn-App: Many non-users
429 were not aware that that it uses Bluetooth technology (Figure7.B). The same pattern of
430 results was observed for Australia’s COVIDSafe app (Garrett, White, et al., 2021).
431 Trust: Our results show that trust in the security of the Corona-Warn-App app plays a
432 crucial role in its uptake. Trust was especially important for past decisions to download the
433 app, while belief in its effectiveness was especially important for future download intentions
434 (see Figures 10 and 11).The observed importance of trust is in line with the findings of
435 other studies exploring attitudes toward and uptake of digital contact tracing technologies:
436 Studies in the United Kingdom (Lewandowsky et al., 2021), Germany (Munzert et al.,
437 2021), and France (Guillon & Kergall, 2020) all show that trust in the government is
438 correlated with the acceptability and use of digital contact tracing apps. Trust in
439 governments, scientists, and health authorities has also been shown to be a strong predictor
440 of adherence to protective measures and guidelines during the pandemic (e.g., Dohle et al.,
441 2020; Han et al., 2021; Udow-Phillips and Lantz, 2020). Likewise, trust has proved to be a COVID-19 TRACKING: GERMANY 26
442 strong predictor of people’s attitudes to other novel technologies, such as 5G (Frey, in
443 press) or gene technology (Siegrist, 2000). These converging findings support the notion
444 that trust in relevant authorities can function as a substitute for knowledge about a
445 technology and its risks and benefits (see Siegrist, 2021).
446 Finally, our findings indicate that pro-free market attitudes, as a proxy for conservative
447 political views, play only a limited role in the adoption of the Corona-Warn-App (see the
448 predictor “Libertarianism” in Figure 10 and Figure 11). This suggests that public health
449 measures such as digital contact tracing technology are not, at least not entirely, evaluated
450 through polarized, partisan perspectives. Because the libertarianism–conservatism
451 dimension is most pronounced in the Anglosphere, and arguably less so in Germany, this
452 finding should be treated with caution. Yet U.S. respondents’ willingness to adopt contact
453 tracing technologies has also been shown to be unrelated to their political leanings, with
454 both Democrats and Republicans almost equally willing to install a potential COVID-19
455 tracking app (Hargittai et al., 2020). Zhang et al. (2020) also found comparable levels of
456 support for government efforts to encourage digital contact tracing among Democrats and
457 Republicans. A similar lack of political polarization was also observed in another relatively
458 novel technological domain, namely, people’s attitudes to personalization of online content
459 (Kozyreva et al., 2021).
460 A Behavioral Framework for Digital Contact Tracing
461 Our analyses highlight several factors that seem to shape people’s attitudes toward
462 digital contact tracing technologies, including privacy concerns, trust in the app’s security,
463 and beliefs about its effectiveness. In order to conceptualize the confluence and interplay of
464 these factors, we suggest mapping them out within a behavior change framework (Michie
465 et al., 2011) such as that shown in Figure 12. This framework consists of three components,
466 whose interaction determines behavior: capability (an individual’s psychological and
467 physical capacity to engage in a behavior), opportunity (environmental affordances and COVID-19 TRACKING: GERMANY 27
468 external factors that enable or prompt a behavior), and motivation (mental processes that
469 guide behavior; e.g., habits, emotions, decisions; Michie et al., 2011).
Figure 12: Behavioral framework for digital contact tracing. Adapted from the behavior change wheel (Michie et al., 2011).
470 Capability encompasses technical capacity (i.e., having a smartphone) and the skills
471 required to download and use the app, as well as the digital skills and risk literacy
472 necessary to understand risk warnings in the app and to communicate test results to the
473 app. In our samples, the majority of participants had a smartphone (Table1), and only
474 about 5% of participants cited not having a smartphone as a reason for not downloading
475 the app (see Figure8). Nevertheless, technical problems related to smartphones (e.g., not
476 having one or the app not working properly) played a prominent role in the open-response
477 questions (see Figure9). Almost all respondents who reported having downloaded the app
478 also reported that it was still installed on their phone (Wave 3: 92%, Wave 4: 93%) and
479 that they kept Bluetooth switched on either always or when leaving the house (Wave 3: COVID-19 TRACKING: GERMANY 28
480 95%, Wave 4: 93%; Appendix Table A5).
481 Opportunity encompasses social and technical/physical factors external to the
482 individual themselves. Social factors include successful communication of the app’s benefits
483 and how to use it, as well as risk communication that explains the risk warnings and their
484 implications for individual behavior. Technical/physical factors include the app’s
485 architecture (e.g., where data are stored, the system’s security) and the broader system in
486 which it is embedded—for example, the health care system. Connecting opportunity in this
487 behavioral framework to the factors that contribute to effectiveness presented in Figure1,
488 it is clear that a digital contact tracing app must be integrated into the national health
489 care system in order to ensure ease of use (e.g., communicating a positive test result
490 anonymously and without friction).
491 Decentralized privacy-respecting apps like the Corona-Warn-App represent a laudable
492 attempt to create an opportunity to contain virus spread that respects the data
493 minimization and protection principles set out in Article 5 of the European Union’s
494 General Data Protection Regulation (European Parliament, 2016). Yet clear
495 communication of the app’s privacy model and risk model is also necessary. Given that
496 most respondents who had not yet downloaded the app did not understand how it works
497 (Figure7.B), it is possible that poorly informed decision making or a knowledge gap keeps
498 uptake unnecessarily low.
499 Motivation encompasses two key factors that are supported by our analyses of the
500 reasons for and predictors of downloading the Corona-Warn-App. One factor comprises
501 people’s intentions to protect themselves and others, to stay informed, and to curb the
502 spread of the virus (Figures8 and9). The other comprises people’s underlying dispositions,
503 such as privacy attitudes, trust in government and technology, and perceptions of risks and
504 benefits (Figures6 and 10). Balance between these two factors is important. For instance,
505 even people driven by prosocial motives may decide against downloading a technology they
506 do not trust. When people’s intentions conflict with their underlying dispositions, the COVID-19 TRACKING: GERMANY 29
507 resulting trade-offs may play into their decision to not adopt digital contact tracing.
508 Taking into account the interdependency of all these factors in a behavior system is
509 essential not only to understanding people’s behaviors regarding digital contact tracing,
510 but also to designing successful behavioral interventions and communication strategies.
511 Conclusion and Policy Implications
512 Which insights from our study could be used to inform public policy? First, it is vital
513 not to compromise on privacy. Although privacy-encroaching measures might initially be
514 accepted in times of crisis, they are unlikely to be accepted on the long term. Moreover,
515 trust in the app’s security was the leading predictor of Corona-Warn-App uptake and data
516 privacy concerns were among the most-cited reasons for not downloading the app. Second,
517 educate people who have not yet downloaded the app about the underlying technology,
518 privacy model, and risk model. Third, make the app and the uploading of test results as
519 simple and convenient as possible. Fourth, address the issue of trust—for example, by
520 effectively demonstrating and communicating how the app preserves privacy, and clarifying
521 that neither the government nor any other institution has access to users’ data.
522 To address effectiveness concerns and increase the benefits for users, more useful
523 functions should be incorporated into digital contact tracing apps. For instance, the
524 Corona-Warn-App now offers information about infection numbers in Germany, a digital
525 check-in system (e.g., for shops or events) via QR codes, and the possibility to integrate
526 digital vaccination certificates (https://www.coronawarn.app/en/blog/).
527 Finally, our findings suggest that arguments for digital contact tracing technologies may
528 be particularly effective when the messaging focuses on prosocial motives, such as helping
529 to stop the spread of the virus, and personal benefits, such as protecting one’s own health.
530 Messaging should also address concerns about the app’s effectiveness and data security.
531 The effectiveness of framing messages along these lines should be empirically tested.
532 If digital contact tracing technologies are to become a long-term solution for managing COVID-19 TRACKING: GERMANY 30
533 viral infectious diseases such as COVID-19, they must be effective, understandable, and
534 acceptable to most people.
535 Methods
536 Participants and Procedure
537 Four representative online samples of German participants (total retained participants
538 N = 4, 357) were recruited through the online platform Lucid using quota sampling to
539 account for current population distributions with regard to age (> 18 years), gender, and
540 region (see Table1 for information about the study, smartphone use, and basic
541 demographics, and Figure2 for data collection times in relation to the pandemic’s
542 development in Germany). Appendix Table A10 provides additional information on
543 educational and regional distribution for the four waves. Informed consent was obtained
544 from all participants and the studies were conducted in accordance with relevant guidelines
545 and regulations. The Institutional Review Board of the Max Planck Institute for Human
546 Development approved the surveys (approval L2020-4).
Table 1 Study and Demographic Information
Wave 1 Wave 2 Wave 3 Wave 4
Recruitment Date of data collection 30–31.03.20 17–22.04.20 25.08–03.09.20 02–08.11.20 Sample size (recruited) 1,224 1,665 1,633 1,518 Sample size (retained) 829 1,109 1,231 1,188 Scenarios Severe, Mild Severe, Mild, Corona-Warn- Corona-Warn- Bluetooth App App Smartphone use (%) No — 3.6 7.4 6.7 Yes — 96.4 92.6 93.3 Gender (%) Female 50.4 50.2 49.6 50.6 Male 49.2 49.3 50.1 49.3 Other 0.4 0.5 0.2 0.1 Age (in years) Median 48.0 48.0 51.0 50.0 SD 17.0 16.0 17.0 18.0 COVID-19 TRACKING: GERMANY 31
547 Study Design
548 There were four waves of data collection (for dates and sample information, see Table1
549 and Figure2), the timing of which was determined by two main criteria: The first criterion
550 was the development of digital contact tracing technology in Germany and worldwide. The
551 project started early in the pandemic, in March 2020, when mobile tracking apps were still
552 at the development stage and public authorities were considering which technology to use
553 (e.g., centralized vs. decentralized). After Germany launched the Corona-Warn-App in
554 June 2020, we replaced the hypothetical scenarios by questions on using the app itself (see
555 Appendix Table B4 for details.)
556 The second criterion was the development of the pandemic and changing infection
557 numbers in Germany. The first sample was collected during the peak of the first wave of
558 infections; the second, when infection rates were going down. Study Waves 3 and 4,
559 focusing on the Corona-Warn-App, were likewise conducted at points with contrasting
560 infection rates: in late summer 2020, when infections were low, and in November 2020,
561 when they were rising steeply. We thus conducted two assessments when infections were
562 peaking (Waves 1 and 4) and two when they were decreasing (Wave 2) or had been low for
563 some time (Wave 3). For a visual representation, see Figure2.
564 There were notable differences in the content of the four assessments (see Figure 13 for
565 a schematic representation), reflecting the ongoing developments in digital contact tracing
566 technology. All surveys shared the same basic structure. Participants first completed an
567 inventory of measures tapping the perceived impact and risks of COVID-19. They then
568 saw one tracking scenario, followed by an inventory of measures tapping their attitudes
569 toward the tracking technologies involved in the scenario seen. This inventory was the
570 same across all scenarios, with one exception: Participants in Waves 1 and 2 were asked
571 about a hypothetical app, whereas participants in Waves 3 and 4 were asked about the
572 Corona-Warn-App itself. All participants also completed an attention check; those who
573 failed to correctly identify the scenario they had seen from three alternatives were excluded COVID-19 TRACKING: GERMANY 32
574 from the analysis. The surveys concluded by assessing respondents’ political worldviews
575 and their attitudes to hypothetical privacy-encroaching measures.
WAVE WAVE WAVE WAVE
Consent, Instructions, Demographics Consent, Instructions, Demographics Consent, Instructions, Demographics Consent, Instructions, Demographics
Impact of COVID-19 Impact of COVID-19 Impact of COVID-19 Impact of COVID-19
Perceived risks of COVID-19 Perceived risks of COVID-19 Perceived risks of COVID-19 Perceived risks of COVID-19
Information source use Information source use & trust Information source use & trust Information source use & trust
Fatalities, & compliance Mild scenario Severe scenario Government response, fatalities, & Government response & compliance compliance Mild Bluetooth Severe Attention check Corona-Warn-App scenario scenario scenario scenario Corona-Warn-App scenario
1st acceptability judgment Attention check Attention check Attention check
Would you download Is this use of data & use this app? acceptable? 1st acceptability judgment Corona-Warn-App downloaded? Corona-Warn-App downloaded?
Would you Is this use of Effectiveness and risk judgments Would you YES NO YES NO download & use this app? data use this app? acceptable? When? Will download When? Will download 2nd acceptability judgment Effectiveness and risk judgments in the future? in the future? Still installed? If no, reasons Still installed? If no, reasons if no is chosen why not why not (open + 2nd acceptability judgment yes no yes no multiple choice) Follow up: Delete data after 6 Bluetooth Why Bluetooth Why months if no is chosen on? removed? on? removed? mild severe Follow up: Delete data after 6 Tried to convince Tried to convince Follow up: Local data Follow up: Opt out months others? others? mild severe Reasons to Reasons to Worldviews: Libertarianism Follow up: Local data Follow up: Opt out download download (open + multiple choice) Upload results if Attitudes toward privacy-encroaching Immunity passports infected? Upload results if measures infected? Estimated public uptake & should the yes no Worldviews: Libertarianism app be mandatory Why? Why not? Attitudes toward privacy-encroaching Effectiveness and risk judgments measures Understanding of Attitudes toward privacy-encroaching risk warnings measures Worldviews: Libertarianism, trust in Estimated public uptake & should the science & tech, conspiracism app be mandatory
Effectiveness and risk judgments
Attitudes toward privacy-encroaching measures Worldviews: Libertarianism, trust in science & tech, conspiracism
Figure 13: Design of Waves 1–4 of the survey. Each box represents a block of one or more questions. Blocks in deeper shades were common elements across all four waves. For the verbatim text of the scenarios (translated from the German), see Appendix Table B4. Full questionnaires (in German) are available at https://osf.io/xvzph.
576 The details of these building blocks—perceived risks, scenario, attitudes toward the
577 scenario presented, attention check, and worldview—differed somewhat between waves.
578 Starting in Wave 2, we also included questions tapping participants’ assessment of the
579 government’s response, estimation of fatalities, and compliance with social distancing rules.
580 From Wave 3, we added further worldview items tapping, for example, attitudes to science COVID-19 TRACKING: GERMANY 33
581 and technology and belief in conspiracy narratives. The scenarios also differed between
582 waves. In Waves 1 and 2, participants were randomly assigned to one of two (Wave 1) or
583 three (Wave 2) hypothetical scenarios; in Waves 3 and 4, they saw a description of the
584 Corona-Warn-App.
585 We now present individual blocks in more detail.
586 Impact of COVID-19 : After providing consent and demographic information,
587 respondents were asked about the impact of COVID-19 on themselves and people they
588 know (see impact items in Appendix Table B1).
589 Perceived risks of COVID-19 : Respondents were then asked about how they perceived
590 the risk of COVID-19 for themselves, other people, and the country as a whole (Risk items
591 summarized in Appendix Table B2).
592 Information source use and trust: They then were asked about which sources they used
593 to inform themselves about the COVID-19 pandemic and how much they trusted these
594 sources (see impact items in Appendix Table B1).
595 Government response, fatalities, and compliance: Respondents then were asked to
596 evaluate government response to the pandemic (Waves 3 and 4), to estimate national
597 fatalities (Waves 2 and 3) and their own policy compliance (Waves 2–4) (see items in
598 Appendix Table B3).
599 Then they were randomly assigned to read a single tracking scenario description (in
600 Waves 1 and 2) or proceeded to read the Corona-Warn-App description (in Waves 3 and 4).
601 Scenarios: The verbatim text of the scenarios (translated from the German) is provided
602 in full in Appendix Table B4. Here, we summarize the main differences.
603 Severe Scenario: In this hypothetical scenario (Waves 1 and 2), anyone with a mobile
604 phone would be tracked; there would be no possibility of opting out. The government could
605 use the data to locate people who violate lockdown orders and fine or arrest them where
606 necessary. Data would also be used to help shape the public health response, to contact
607 people potentially exposed to COVID-19, and to issue individual quarantine orders. COVID-19 TRACKING: GERMANY 34
608 Mild Scenario: In this hypothetical scenario (Waves 1 and 2), only people who
609 download a government app and agree to be tracked and contacted would be included.
610 Data would only be used to contact those potentially exposed to COVID-19.
611 Bluetooth Scenario: In this hypothetical scenario (Wave 2 only), respondents were
612 informed about the Bluetooth-based contact tracing function proposed by Apple and
613 Google. People potentially exposed to COVID-19 would be notified without the
614 government knowing who they are. The use of this contact tracing capability would be
615 completely voluntary. Those notified would not know the identity of the person who had
616 tested positive.
617 Corona-Warn-App: In this scenario (Waves 3 and 4), respondents were informed about
618 the Bluetooth-based app launched in Germany on 16 June 2020. Users who test positive
619 for COVID-19 are asked whether they want to share their result for contact tracing. The
620 app does not evaluate any geodata and does not transmit any location information. No
621 personal data is sent or stored. The anonymized contact data is stored locally on the user’s
622 smartphone.
623 Acceptability and uptake: In Waves 1 and 2, participants answered a series of questions
624 assessing the acceptability of the scenario they had viewed, as well as their willingness to
625 use the app described (for items, see Appendix Table B5. Binary acceptability judgments
626 (“Would you download and use the app?” for the mild and Bluetooth scenarios, and “Is
627 this use of the tracking data acceptable?” for the severe scenario) were tapped twice:
628 immediately after participants read the scenario and again after they had answered
629 questions (standardized across waves) about the effectiveness and risks of the app
630 presented. Participants who answered “No” after the second acceptability judgment were
631 then asked follow-up questions probing whether their decision would change if the
632 government (or Google and Apple in the Bluetooth scenario) were obliged to delete all data
633 and stop tracking after 6 months. In the severe scenario, participants were also asked
634 whether their decision would change if they had the option to opt out of data collection; in COVID-19 TRACKING: GERMANY 35
635 the mild scenario, they were asked if their decision would change if their data were only
636 stored locally (see Appendix Table B5 for all questions). In Waves 3 and 4, participants
637 read the Corona-Warn-App scenario and were then asked whether they had downloaded
638 the app or planned to download it in the future; they were also asked about their app
639 usage and about the reasons for downloading/not downloading the app or uploading/not
640 uploading their test results (multiple-choice format; multiple responses to this questions
641 were allowed) (see Appendix Table B5). In Wave 4, we asked participants to describe their
642 reasons in their own words (open-response format) before presenting them with the same
643 multiple-choice questions (see Appendix Table B7).
644 Perceived effectiveness and risks of tracking technologies: Next block of questions
645 probed respondents’ perceptions of the app’s effectiveness and potential risks (see
646 Appendix Table B6 for the full list of questions).
647 Worldviews: We collected information about participants’ worldviews, including
648 attitudes toward the free market (based on Heath and Gifford, 2006; Lewandowsky et al.,
649 2013); higher mean responses reflected more conservative/libertarian worldviews. In Waves
650 3 and 4, we also surveyed respondents’ trust in science and endorsement of conspiracy
651 beliefs. To measure conspiracy beliefs in Wave 3, we adapted a general conspiracy scale
652 from Imhoff and Bruder (2014), selecting the five items with the highest item–total
653 correlations and adding one additional item specifically tailored to the COVID-19 pandemic
654 (“Selfish interests have conspired to convince the public that COVID-19 is a major threat,”
655 designed based on the conspiracy beliefs inventory by van der Linden et al., 2021). In Wave
656 4, we created our own items based on COVID-19-related conspiracy narratives that were
657 growing in popularity at the time. To counteract this exposure to conspiracy narratives, we
658 included a debriefing flyer developed on the basis of European Commission (2020)
659 recommendations at the end of the survey. For all worldview items, see Appendix Table B8.
660 Attitudes toward privacy-encroaching measures: Finally, we asked respondents to rate
661 the acceptability of six measures that could limit the spread of COVID-19. These COVID-19 TRACKING: GERMANY 36
662 hypothetical measures, which could potentially compromise people’s privacy, included
663 giving the government access to medical records, tracking people’s locations using mobile
664 phone data, and temporarily relaxing data protection regulations (for a full list, see
665 Appendix Table B9).
666 Data Analysis and Reporting
667 We used logistic regression analyses to identify predictors of Corona-Warn-App
668 downloads in Waves 3 and 4 of the survey. All combined measures used in the regressions
669 analyses are summarized in Table B10. For regression analyses, gender coefficient was
670 assessed as a binary variable, and "Other" category was excluded from regression analyses
671 due to its small sample size (see Table1). To analyze the open-response question on why
672 people did or did not download the app, we counted the frequencies with which terms
673 occurred across different respondents’ responses and the frequency with which the terms
674 co-occurred within the same respondent’s response. Based on these frequencies we built
675 co-occurrence networks of unigrams (i.e., individual words) using a simple feature
676 extraction method from the Python package scikit-learn (version 0.24.1) to collect unigram
677 frequencies (Pedregosa et al., 2011); we used the graph-tool library (version 2.37) to build
678 networks of unigrams according to their co-occurrences within a response (Peixoto, 2014).
679 In this article, we report selected results relevant for understanding public attitudes toward
680 privacy and tracking technologies during the pandemic. Full descriptive results for all four
681 waves of the survey are available online: https://ai_society.mpib.dev/tracking-app.
682 Data availability
683 Anonymized data and code are available at (https://osf.io/xvzph).
684 Supporting information
685 AppendixA contains additional figures and tables to support our reporting. Appendix
686 B contains tables listing the items administered and covariates for our models. The COVID-19 TRACKING: GERMANY 37
687 German version of the study questionnaires is available at https://osf.io/xvzph.
688 Funding and acknowledgements
689 The study was funded by a Volkswagen Foundation grant to R.H., S.L., and S.M.H.
690 (Initiative “Artificial Intelligence and the Society of the Future”). The study is part of an
691 international initiative and we are grateful to our partners at the University of Melbourne
692 Complex Human Data Hub, led by Simon Dennis and supported by Andrew Perfors,
693 Yoshihisa Kashima, Joshua White, and Daniel Little. The authors thank the Volkswagen
694 Foundation for providing financial support and Lucid for their help with data collection.
695 We thank Deb Ain and Susannah Goss for editing the manuscript, Marlene Wulf and
696 Larissa Samaan for research assistance, and our colleagues at the Center for Adaptive
697 Rationality for their feedback and productive discussions.
698 Authors’ contributions: CRediT
699 Conceptualization: S.L., P.M.G., S.M.H., A.K., P.L.-S., T.P., and R.H. Data curation:
700 A.K. and P.L.-S. Formal analysis: A.K., P.L.-S., P.M.G., and S.M.H. Funding acquisition:
701 P.L.-S., S.L., S.M.H., and R.H. Investigation: A.K. and P.L.-S. Methodology: S.L., P.M.G.,
702 S.M.H., A.K., P.L.-S., T.P., and R.H. Project administration: A.K. and P.L.-S. Supervision:
703 S.L., S.M.H., and R.H. Visualization: A.K., P.L.-S., S.M.H., and P.M.G. Writing - original
704 draft: A.K. Writing - review & editing: A.K., P.L.-S., S.L., P.M.G., S.M.H., T.P., and R.H.
705 Correspondence concerning this article should be addressed to Anastasia Kozyreva,
706 Center for Adaptive Rationality, Max Planck Institute for Human Development, Berlin.
707 Email: [email protected] COVID-19 TRACKING: GERMANY 38
708 References
709 Aleta, A., Martín-Corral, D., Pastore y Piontti, A., Ajelli, M., Litvinova, M., Chinazzi, M.,
710 Dean, N. E., Halloran, M. E., Longini Jr., I. M., Merler, S., Pentland, A.,
711 Vespignani, A., Moro, E., & Moreno, Y. (2020). Modelling the impact of testing,
712 contact tracing and household quarantine on second waves of COVID-19. Nature
713 Human Behaviour, 4 (9), 964–971. https://doi.org/10.1038/s41562-020-0931-9
714 Bastian, M., Heymann, S., & Jacomy, M. (2009). Gephi: An open source software for
715 exploring and manipulating networks.
716 http://www.aaai.org/ocs/index.php/ICWSM/09/paper/view/154
717 Bianconi, G., Sun, H., Rapisardi, G., & Arenas, A. (2021). Message-passing approach to
718 epidemic tracing and mitigation with apps. Physical Review Research, 3 (1), Article
719 L012014. https://doi.org/10.1103/PhysRevResearch.3.L012014
720 Burdinski, A., Brockmann, D., & Maier, B. F. (2021). Digital contact tracing contributes
721 little to covid-19 outbreak containment. medRxiv.
722 https://doi.org/10.1101/2021.06.21.21259258
723 Cho, H., Ippolito, D., & Yu, Y. W. (2020). Contact tracing mobile apps for covid-19:
724 Privacy considerations and related trade-offs. https://arxiv.org/abs/2003.11511
725 Colizza, V., Grill, E., Mikolajczyk, R., Cattuto, C., Kucharski, A., Riley, S., Kendall, M.,
726 Lythgoe, K., Bonsall, D., Wymant, C., Abeler-Dörner, L., Ferretti, L., & Fraser, C.
727 (2021). Time to evaluate COVID-19 contact-tracing apps. Nature Medicine, 27 (3),
728 361–362. https://doi.org/10.1038/s41591-021-01236-6
729 Danquah, L. O., Hasham, N., MacFarlane, M., Conteh, F. E., Momoh, F., Tedesco, A. A.,
730 Jambai, A., Ross, D. A., & Weiss, H. A. (2019). Use of a mobile application for
731 Ebola contact tracing and monitoring in northern Sierra Leone: A proof-of-concept
732 study. BMC Infectious Diseases, 19 (1), Article 810.
733 https://doi.org/10.1186/s12879-019-4354-z COVID-19 TRACKING: GERMANY 39
734 Dienlin, T., & Metzger, M. J. (2016). An extended privacy calculus model for SNSs:
735 Analyzing self-disclosure and self-withdrawal in a representative U.S. sample.
736 Journal of Computer-Mediated Communication, 21 (5), 368–383.
737 https://doi.org/10.1111/jcc4.12163
738 Dohle, S., Wingen, T., & Schreiber, M. (2020). Acceptance and adoption of protective
739 measures during the covid-19 pandemic: The role of trust in politics and trust in
740 science. Social Psychological Bulletin, 15 (4), 1–23.
741 https://doi.org/10.32872/spb.4315
742 European Commission. (2020). So erkennt man Verschwörungstheorien.
743 https://ec.europa.eu/info/live-work-travel-eu/coronavirus-response/fighting-
744 disinformation/identifying-conspiracy-theories_de
745 European Parliament. (2016). Regulation (EU) 2016/679 of the European Parliament and
746 of the Council of 27 April 2016 on the protection of natural persons with regard to
747 the processing of personal data and on the free movement of such data, and
748 repealing directive 95/46/EC (General Data Protection Regulation).
749 http://data.europa.eu/eli/reg/2016/679/oj
750 Ferretti, L., Wymant, C., Kendall, M., Zhao, L., Nurtay, A., Abeler-Dörner, L., Parker, M.,
751 Bonsall, D., & Fraser, C. (2020). Quantifying SARS-CoV-2 transmission suggests
752 epidemic control with digital contact tracing. Science, 368 (6491), Article eabb6936.
753 https://doi.org/10.1126/science.abb6936
754 Fischhoff, B., Slovic, P., Lichtenstein, S., Read, S., & Combs, B. (1978). How safe is safe
755 enough? a psychometric study of attitudes towards technological risks and benefits.
756 Policy sciences, 9 (2), 127–152.
757 Frey, R. (in press). Psychological drivers of individual differences in risk perception: A
758 systematic case study focusing on 5g. Psychological Science.
759 https://doi.org/10.31234/osf.io/7a3kg COVID-19 TRACKING: GERMANY 40
760 Garrett, P. M., Wang, Y., White, J. P., Hsieh, S., Strong, C., Lee, Y.-C., Lewandowsky, S.,
761 Dennis, S., & Yang, C.-T. (2021). Young adults view smartphone tracking
762 technologies for COVID-19 as acceptable: The case of Taiwan. International Journal
763 of Environmental Research and Public Health, 18 (3), Article 1332.
764 https://doi.org/10.3390/ijerph18031332
765 Garrett, P. M., White, J. P., Lewandowsky, S., Kashima, Y., Perfors, A., Little, D. R.,
766 Geard, N., Mitchell, L., Tomko, M., & Dennis, S. (2021). The acceptability and
767 uptake of smartphone tracking for COVID-19 in Australia. PLOS ONE, 16 (1),
768 Article e0244827. https://doi.org/10.1371/journal.pone.0244827
769 Gelman, A. (2008). Scaling regression inputs by dividing by two standard deviations.
770 Statistics in Medicine, 27 (15), 2865–2873. https://doi.org/10.1002/sim.3107
771 Grantz, K. H., Meredith, H. R., Cummings, D. A. T., Metcalf, C. J. E., Grenfell, B. T.,
772 Giles, J. R., Mehta, S., Solomon, S., Labrique, A., Kishore, N., Buckee, C. O., &
773 Wesolowski, A. (2020). The use of mobile phone data to inform analysis of
774 COVID-19 pandemic epidemiology. Nature Communications, 11 (1), Article 4961.
775 https://doi.org/10.1038/s41467-020-18190-5
776 Gstraunthaler, T., & Day, R. (2008). Avian influenza in the uk: Knowledge, risk perception
777 and risk reduction strategies. British Food Journal, 110 (3), 260–270.
778 https://doi.org/10.1108/00070700810858673
779 Guillon, M., & Kergall, P. (2020). Attitudes and opinions on quarantine and support for a
780 contact-tracing application in France during the COVID-19 outbreak. Public Health,
781 188, 21–31. https://doi.org/10.1016/j.puhe.2020.08.026
782 Habersaat, K. B., Betsch, C., Danchin, M., Sunstein, C. R., Böhm, R., Falk, A.,
783 Brewer, N. T., Omer, S. B., Scherzer, M., Sah, S., Fischer, E. F., Scheel, A. E.,
784 Fancourt, D., Kitayama, S., Dubé, E., Leask, J., Dutta, M., MacDonald, N. E.,
785 Temkina, A., . . . Butler, R. (2020). Ten considerations for effectively managing the COVID-19 TRACKING: GERMANY 41
786 COVID-19 transition. Nature Human Behaviour, 4 (7), 677–687.
787 https://doi.org/10.1038/s41562-020-0906-x
788 Han, Q., Zheng, B., Cristea, M., Agostini, M., Bélanger, J. J., Gützkow, B.,
789 Kreienkamp, J., & Leander, N. P. (2021). Trust in government regarding covid-19
790 and its associations with preventive health behaviour and prosocial behaviour
791 during the pandemic: A cross-sectional and longitudinal study. Psychological
792 Medicine, 1–11. https://doi.org/10.1017/S0033291721001306
793 Hargittai, E., Redmiles, E. M., Vitak, J., & Zimmer, M. (2020). Americans’ willingness to
794 adopt a covid-19 tracking app. First Monday.
795 https://doi.org/10.5210/fm.v25i11.11095
796 Hart, V., Siddarth, D., Cantrell, B., Tretikov, L., Eckersley, P., Langford, J., Leibrand, S.,
797 Kakade, S., Latta, S., Lewis, D., Tessaro, S., & Weyl, G. (2020). Outpacing the
798 virus: Digital response to containing the spread of covid-19 while mitigating privacy
799 risks. https://ethics.harvard.edu/outpacing-virus
800 Heath, Y., & Gifford, R. (2006). Free-market ideology and environmental degradation: The
801 case of belief in global climate change. Environment and Behavior, 38 (1), 48–71.
802 https://doi.org/10.1177/0013916505277998
803 Hinch, R., Probert, W., Nurtay, A., Kendall, M., Wymant, C., Hall, M., Lythgoe, K.,
804 Bulas Cruz, A., Zhao, L., Stewart, A., Ferretti, L., Parker, M., Montero, D.,
805 Warren, J., Mather, N. K., Finkelstein, A., Abeler-Dörner, L., Bonsall, D., &
806 Fraser, C. (2020). Effective configurations of a digital contact tracing app: A report
807 to NHSX [Accessed: 2021-02-09]. https://github.com/BDI-pathogens/covid-
808 19_instant_tracing/blob/master/Report%20-%20Effective%20Configurations%
809 20of%20a%20Digital%20Contact%20Tracing%20App.pdf
810 Imhoff, R., & Bruder, M. (2014). Speaking (un-)truth to power: Conspiracy mentality as a
811 generalised political attitude. European Journal of Personality, 28 (1), 25–43.
812 https://doi.org/10.1002/per.1930 COVID-19 TRACKING: GERMANY 42
813 Kahn, J. P., & Johns Hopkins Project on Ethics and Governance of Digital Contact
814 Tracing Technologies (Eds.). (2020). Digital contact tracing for pandemic response:
815 Ethics and governance guidance. Johns Hopkins University Press.
816 https://doi.org/doi:10.1353/book.75831
817 Kaptchuk, G., Goldstein, D. G., Hargittai, E., Hofman, J., & Redmiles, E. M. (2020). How
818 good is good enough for COVID19 apps? The influence of benefits, accuracy, and
819 privacy on willingness to adopt. arXiv preprint arXiv:2005.04343.
820 Kozyreva, A., Lorenz-Spreen, P., Hertwig, R., Lewandowsky, S., & Herzog, S. (2021).
821 Public attitudes towards algorithmic personalization and use of personal data
822 online: Evidence from Germany, Great Britain, and the US. 8 (117).
823 https://doi.org/10.1057/s41599-021-00787-w
824 Lewandowsky, S., Dennis, S., Perfors, A., Kashima, Y., White, J. P., Garrett, P.,
825 Little, D. R., & Yesilada, M. (2021). Public acceptance of privacy-encroaching
826 policies to address the COVID-19 pandemic in the United Kingdom. PLOS ONE,
827 16 (1), Article e0245740. https://doi.org/10.1371/journal.pone.0245740
828 Lewandowsky, S., Gignac, G. E., & Oberauer, K. (2013). The role of conspiracist ideation
829 and worldviews in predicting rejection of science. PLOS ONE, 8 (10), Article
830 e75637. https://doi.org/10.1371/journal.pone.0075637
831 Michie, S., van Stralen, M. M., & West, R. (2011). The behaviour change wheel: A new
832 method for characterising and designing behaviour change interventions.
833 Implementation Science, 6 (1), Article 42.
834 https://doi.org/https://doi.org/10.1186/1748-5908-6-42
835 Munzert, S., Selb, P., Gohdes, A., Stoetzer, L. F., & Lowe, W. (2021). Tracking and
836 promoting the usage of a COVID-19 contact tracing app. Nature Human Behaviour,
837 5 (2), 247–255. https://doi.org/10.1038/s41562-020-01044-x
838 Oliver, N., Lepri, B., Sterly, H., Lambiotte, R., Deletaille, S., De Nadai, M., Letouzé, E.,
839 Salah, A. A., Benjamins, R., Cattuto, C., Colizza, V., de Cordes, N., COVID-19 TRACKING: GERMANY 43
840 Fraiberger, S. P., Koebe, T., Lehmann, S., Murillo, J., Pentland, A., Pham, P. N.,
841 Pivetta, F., . . . Vinck, P. (2020). Mobile phone data for informing public health
842 actions across the COVID-19 pandemic life cycle. Science Advances, 6 (23), Article
843 eabc0764. https://doi.org/10.1126/sciadv.abc0764
844 O’Neill, P. H., Ryan-Mosley, T., & Johnson, B. (2020). A flood of coronavirus apps are
845 tracking us. now it’s time to keep track of them.
846 https://www.technologyreview.com/2020/05/07/1000961/launching-mittr-covid-
847 tracing-tracker
848 Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O.,
849 Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A.,
850 Cournapeau, D., Brucher, M., Perrot, M., & Duchesnay, E. (2011). Scikit-learn:
851 Machine learning in Python. Journal of Machine Learning Research, 12, 2825–2830.
852 Peixoto, T. P. (2014). The graph-tool python library. figshare.
853 https://doi.org/10.6084/m9.figshare.1164194
854 Redmiles, E. M. (2020). User concerns & tradeoffs in technology-facilitated contact tracing.
855 arXiv preprint arXiv:2004.13219. https://doi.org/10.1145/3428093
856 Robert Koch Institute. (2020). Die Pandemie in Deutschland in den nächsten Monaten -
857 Ziele, Schwerpunktthemen und Instrumente für den Infektionsschutz.
858 Strategie-Ergänzung, Stand 23.10.2020 [The pandemic in Germany in the coming
859 months: Goals, key issues, and tools for infection control. Strategy supplement, as of
860 23.10.2020] (tech. rep.). https://www.rki.de/DE/Content/InfAZ/N/Neuartiges_
861 Coronavirus/Strategie_Ergaenzung_Covid.html
862 Rodríguez, P., Graña, S., Alvarez-León, E. E., Battaglini, M., Darias, F. J., Hernán, M. A.,
863 López, R., Llaneza, P., Martín, M. C., RadarCovidPilot Group, Ramirez-Rubio, O.,
864 Romaní, A., Suárez-Rodríguez, B., Sánchez-Monedero, Arenas, A., & Lacasa, L.
865 (2021). A population-based controlled experiment assessing the epidemiological COVID-19 TRACKING: GERMANY 44
866 impact of digital contact tracing. Nature Communications, 12 (1), Article 587.
867 https://doi.org/10.1038/s41467-020-20817-6
868 Siegrist, M. (2000). The influence of trust and perceptions of risks and benefits on the
869 acceptance of gene technology. Risk analysis, 20 (2), 195–204.
870 Siegrist, M. (2021). Trust and risk perception: A critical review of the literature. Risk
871 analysis, 41 (3), 480–490.
872 Simko, L., Chang, J. L., Jiang, M., Calo, R., Roesner, F., & Kohno, T. (2020). COVID-19
873 contact tracing and privacy: A longitudinal study of public opinion. arXiv preprint
874 arXiv:2012.01553.
875 Slovic, P. (1987). Perception of risk. Science, 236 (4799), 280–285.
876 Udow-Phillips, M., & Lantz, P. M. (2020). Trust in public health is essential amid the
877 covid-19 pandemic. Journal of Hospital Medicine, 15 (7), 431–433.
878 https://doi.org/10.12788/jhm.3474
879 van der Linden, S., Panagopoulos, C., Azevedo, F., & Jost, J. T. (2021). The paranoid
880 style in American politics revisited: An ideological asymmetry in conspiratorial
881 thinking. Political Psychology, 42 (1), 23–51. https://doi.org/10.1111/pops.12681
882 Whitelaw, S., Mamas, M. A., Topol, E., & Van Spall, H. G. C. (2020). Applications of
883 digital technology in COVID-19 pandemic planning and response. The Lancet
884 Digital Health, 2 (8), e435–e440. https://doi.org/10.1016/S2589-7500(20)30142-4
885 World Health Organization. (2020). COVID-19 operationalization of the global response
886 strategy in the WHO European Region: September 2020 (tech. rep.). WHO Regional
887 Office for Europe. https://apps.who.int/iris/bitstream/handle/10665/334167/WHO-
888 EURO-2020-1073-408190-55167-eng.pdf
889 Wymant, C., Ferretti, L., Tsallis, D., Charalambides, M., Abeler-Dörner, L., Bonsall, D.,
890 Hinch, R., Kendall, M., Milsom, L., Ayres, M., et al. (2021). The epidemiological
891 impact of the nhs covid-19 app. Nature, 594 (7863), 408–412.
892 https://doi.org/10.1038/s41586-021-03606-z COVID-19 TRACKING: GERMANY 45
893 Zhang, B., Kreps, S., McMurry, N., & McCain, R. M. (2020). Americans’ perceptions of
894 privacy and surveillance in the covid-19 pandemic. Plos one, 15 (12), e0242652.
895 https://doi.org/10.1371/journal.pone.0242652 COVID-19 TRACKING: GERMANY 46
Appendix A Supplementary Information: Figures and Tables
COVID−19 contact tracing apps: Downloads
Singapore 82
Finland 45
Switzerland 35
Germany 32
UK 31
Australia 28
France 19
Italy 17
Spain 15
India 13
0 25 50 75 100 % Downloads Figure A1: Adoption of COVID-19 contact tracing apps in selected countries, as indicated by the percentage of the population that has downloaded the app. See Table A1 for detailed information. Latest update: April 7, 2021. COVID-19 TRACKING: GERMANY 47
Table A1 Adoption of COVID-19 Contact Tracing Apps in Selected Countries: Further Information on Apps and Download Figures
Country Name Developer/Deployer Technology Release Downloads Downloads Numbers date (N) (%) updated on
Singapore TraceTogether GovTech Agency Bluetooth, 20.03.2020 4,700,000 82% 07.04.2021* BlueTrace Finland Koronavilkku Finnish Institute of Bluetooth, 31.08.2020 2,500,000 45% 05.11.2020** Health and Welfare Google/Apple Switzerland SwissCovid Swiss National Bluetooth, 23.07.2020 3,059,000 35% 06.04.2021 Covid-19 Science Google/Apple Task Force Germany Corona-Warn- Deutsche Telekom, Bluetooth, 16.06.2020 26,700,000 32% 01.04.2021 App SAP/Robert Koch Google/Apple Institute United NHS COVID- NHS Bluetooth, 24.09.2020 20,900,000 31% 23.12.2020** Kingdom 19 App Google/Apple Australia COVIDSafe Australian govern- Bluetooth 26.04.2020 7,000,000 28% 07.04.2020* ment France TousAntiCovid Inria Bluetooth 22.10.2020 13,000,000 19% 01.03.2021** Italy Immuni Bending Spoons Bluetooth, 01.06.2020 10.400.709 17% 01.04.2021 Google/Apple Spain Radar COVID Ministry of Eco- Bluetooth, August 7,200,000 15% 28.03.2021 nomic Affairs and Google/Apple 2020 Digital Transfor- mation India Aarogya Setu Indian national Bluetooth, Lo- 02.04.2020 173,700,000 13% 07.04.2020 government cation * The app’s website provides only approximate numbers of downloads and no information about when the numbers were updated. ** The app’s website provides no information about downloads; the reported numbers are taken from press coverage or statista.com.
Table A2 Cochran–Armitage Trend Test for Figure4
Measure n χ2 p value df
Data on infections & immunity status 4302 24.208 <.001 1 Access to medical records 4302 239.253 <.001 1 Location tracking data 4302 84.693 <.001 1 Notifications of leaving quarantine 4302 16.747 <.001 1 Suspension of data protection 4302 6.674 .01 1 Data on contacts & interactions 4302 5.883 .02 1 COVID-19 TRACKING: GERMANY 48
Table A3 1-Sample Proportions Test with Continuity Correction (prop.test in R) for Figure4
Measure Wave N accept % accept Total N 95% CI (low) 95% CI (high)
Data on contacts & interactions Wave 1 387 49 788 0.46 0.53 Suspension of data protection Wave 1 409 52 788 0.48 0.55 Data on infections & immunity status Wave 1 541 69 788 0.65 0.72 Location tracking data Wave 1 470 60 788 0.56 0.63 Access to medical records Wave 1 533 68 788 0.64 0.71 Notifications of leaving quarantine Wave 1 465 59 788 0.55 0.62
Data on contacts & interactions Wave 2 498 45 1102 0.42 0.48 Suspension of data protection Wave 2 525 48 1102 0.45 0.51 Data on infections & immunity status Wave 2 719 65 1102 0.62 0.68 Location tracking data Wave 2 605 55 1102 0.52 0.58 Access to medical records Wave 2 649 59 1102 0.56 0.62 Notifications of leaving quarantine Wave 2 549 50 1102 0.47 0.53
Data on contacts & interactions Wave 3 453 37 1230 0.34 0.40 Suspension of data protection Wave 3 548 45 1230 0.42 0.47 Data on infections & immunity status Wave 3 669 54 1230 0.52 0.57 Location tracking data Wave 3 576 47 1230 0.44 0.50 Access to medical records Wave 3 590 48 1230 0.45 0.51 Notifications of leaving quarantine Wave 3 637 52 1230 0.49 0.55
Data on contacts & interactions Wave 4 536 45 1182 0.42 0.48 Suspension of data protection Wave 4 547 46 1182 0.43 0.49 Data on infections & immunity status Wave 4 714 60 1182 0.58 0.63 Location tracking data Wave 4 481 41 1182 0.38 0.44 Access to medical records Wave 4 413 35 1182 0.32 0.38 Notifications of leaving quarantine Wave 4 565 48 1182 0.45 0.51
Table A4 1-Sample Proportions Test With Continuity Correction (prop.test in R) for Figure5
Scenario type Wave Item N accept % accept Total N 95% CI (low) 95% CI (high)
Severe Wave 1 Acceptability of the scenario 238 56 425 0.51 0.61 Severe Wave 1 With follow up: delete data 292 69 425 0.64 0.73 Severe Wave 1 With follow up: opt out 334 79 425 0.74 0.82 Severe Wave 2 Acceptability of the scenario 223 55 407 0.50 0.60 Severe Wave 2 With follow up: delete data 280 69 407 0.64 0.73 Severe Wave 2 With follow up: opt out 327 80 407 0.76 0.84
Mild Wave 1 Acceptability of the scenario 245 61 404 0.56 0.65 Mild Wave 1 With follow up: delete data 287 71 404 0.66 0.75 Mild Wave 1 With follow up: local data 300 74 404 0.70 0.78 Mild Wave 2 Acceptability of the scenario 232 64 362 0.59 0.69 Mild Wave 2 With follow up: delete data 259 72 362 0.67 0.76 Mild Wave 2 With follow up: local data 286 79 362 0.74 0.83
Bluetooth Wave 2 Acceptability of the scenario 200 59 340 0.53 0.64 Bluetooth Wave 2 With follow up: delete data 228 67 340 0.62 0.72
Corona-Warn-App Wave 3 Downloads 446 36 1231 0.34 0.39 Corona-Warn-App Wave 3 With follow up: future downloads 634 52 1231 0.49 0.54 Corona-Warn-App Wave 4 Downloads 483 41 1188 0.38 0.44 Corona-Warn-App Wave 4 With follow up: future downloads 657 55 1188 0.52 0.58 COVID-19 TRACKING: GERMANY 49
Concern for self vs. others
100
75 59 59 49 49 50 43 44 33 33 25
% of respondents 8 7 8 8 0
More concern for self Equal concern More concern for others
Wave 1 Wave 2 Wave 3 Wave 4
Figure A2: Concern for self and others: COVID-19 risk perceptions within respondents. Responses are grouped into three categories: (1) respondents who rated concern for themselves higher than concern for others, (2) respondents who gave the same rating to both, and (3) respondents who rated concern for others higher than concern for themselves. Questions: (1) Concern self: How concerned are you that you might become infected with COVID-19? (2) Concern others: How concerned are you that somebody you know might become infected with COVID-19?"
Table A5 Corona-Warn-App Usage
Wave 3 Wave 4
Have you downloaded the Corona-Warn-App? Yes 36.2 40.7 No 63.8 59.3 Is the Corona-Warn-App still installed on your phone? Yes 91.6 93.2 No 8.4 6.8 Do you generally have Bluetooth switched on so the Corona-Warn-App can operate effectively? Yes 76.9 74.4 No 3.3 4.9 Only when I leave the house 18.0 18.4 I don’t know 1.8 2.2 Have you made any attempts to convince your friends and/or family to download the Corona-Warn-App? Yes 73.1 67.3 No 26.9 32.7 Will you download the Corona-Warn-App in the future? Yes 23.9 24.7 No 76.1 75.3 Do you think that the government should make the Corona-Warn-App mandatory? Yes 28.8 30.1 No 71.2 69.9 COVID-19 TRACKING: GERMANY 50
Acceptability of measures and COVID−19 risk perception
Data on infections & Access to immunity status medical records 2 R = 0.24, p < 0.001 R = 0.15, p < 0.001 R = 0.18, p < 0.001 R = 0.099, p = 0.001 1 R = 0.22, p < 0.001 R = 0.18, p < 0.001 R = 0.29, p < 0.001 R = 0.23, p < 0.001
0
−1
−2 −1 0 1 2 −2 −1 0 1 2
Location Notifications of tracking data leaving quarantine 2 R = 0.31, p < 0.001 R = 0.26, p < 0.001 R = 0.2, p < 0.001 R = 0.18, p < 0.001 1 R = 0.22, p < 0.001 R = 0.26, p < 0.001 R = 0.24, p < 0.001 R = 0.31, p < 0.001
0
−1
−2 −1 0 1 2 −2 −1 0 1 2 Privacy−encroachement acceptance level Privacy−encroachement Suspension of Data on contacts data protection & interactions 2 R = 0.24, p < 0.001 R = 0.28, p < 0.001 R = 0.12, p < 0.001 R = 0.18, p < 0.001 1 R = 0.19, p < 0.001 R = 0.2, p < 0.001 R = 0.24, p < 0.001 R = 0.27, p < 0.001
0
−1
−2 −1 0 1 2 −2 −1 0 1 2 Risk perception
a Wave 1 a Wave 2 a Wave 3 a Wave 4
Figure A3: Correlations between acceptability of various privacy-encroaching measures and COVID-19 risk perceptions within respondents across all four waves of the survey. The risk perception variable is a combined score for the four items in Figure3. All variables are center-scaled. Individual responses are jittered. Lines represent simple linear regression slopes and their 95% confidence band. COVID-19 TRACKING: GERMANY 51
Comparing risks and benefits of CWA within respondents
Belief in effectiveness Trust in security Trust in security vs. risk of harm vs. risk of harm vs. belief in effectiveness
100 CWA downloaded: Yes 86 82 85 82 77 75 70
50 30 25 23 14 18 15 18
0
100 CWA downloaded: No
75 % of respondents 65 62 61 60 51 50 50 49 50 40 38 39 35 25
0
Lower or equal Higher Lower or equal Higher Lower or equal Higher
wave 3 wave 4
Figure A4: Within-participants comparisons of perceived effectiveness and risks of the Corona-Warn-App. The variables “Trust in security” and “Belief in effectiveness” represent combined and averaged measures from variables presented in Figure6. The variable “Trust in security” combines four variables: “Trust: Data for Pandemic Only,” “Trust: Privacy Protection,” “Only Necessary Data Collected” “Trust: Security from 3rd Party.” The variable “Belief in effectiveness” combines three variables: “Reduce Likelihood to Contract,” “Reduce Spread Future,” “Return to Activity Future.” “Risk of harm” is a single variable. For items, see Appendix Tables B6. The figure complements Figure7A and shows comparisons on the individual level. COVID-19 TRACKING: GERMANY 52
Corona−Warn−App downloads by gender
male female 100
75 68 60 62 413 56 370 374 330 50 44 256 40 38 247 227 32 198 25 % of respondents
0
Yes No Yes No
Corona−Warn−App downloads by education
low medium high 100
75 71 479 65 406 56 54 54 49 162 51 144 140 50 151 159 46 46 44 121 118 35 127 29 214 25 198 % of respondents
0
Yes No Yes No Yes No
Corona−Warn−App downloads by age 100
75
Age 50
25
Yes No
Wave 3 Wave 4
Figure A5: Reported Corona-Warn-App downloads by demographic characteristics. For education, “low” comprises the categories “Realschule,” “Hauptschule,” and “None” as the highest level of education; “medium” indicates “Abitur;” and “high” indicates “University”. For gender, the "Other" category was excluded from this analysis due to its small sample size (see Table1). Barplots: Percentages shown in black; respondent numbers shown in white. Boxplots: Boxes show the interquartile range (IQR) of the age distribution (values between the 25th and 75th percentiles); the black lines inside the boxes indicate the median value. Lower and upper whiskers extend from the hinge to the smallest and largest values within 1.5 × IQR. Individual responses are jittered both horizontally and vertically. COVID-19 TRACKING: GERMANY 53
Correlation matrix: Wave 3
Gender_male
Government_announcements_use_and_trust 0
Social_media_use_and_trust 0.1 −0.1
Approval_of_German_pandemic_response 0 0.5 0.1
Belief_in_CWA_effectiveness 0.4 0.2 0.4 0
Age −0.1 0.1 −0.2 0 0.2
CWA_risk_of_harm_perception −0.1 −0.2 −0.3 0.2 −0.3 −0.1 1.0
0.5 Trust_in_CWA_security −0.5 0.1 0.6 0.4 0 0.5 0.1 0.0
Find_government_guidelines_helpful 0.5 −0.3 0.1 0.5 0.6 0 0.6 0 −0.5
−1.0 COVID19_risk_perception 0.4 0.3 −0.1 0.1 0.3 0.2 0.1 0.4 −0.1
Attitudes_to_science_and_technology 0.2 0.3 0.4 −0.3 0.1 0.3 0.3 0 0.3 0.1
Belief_in_conspiracies −0.2 −0.2 −0.4 −0.5 0.4 −0.1 −0.2 −0.3 0.2 −0.4 −0.1
Libertarianism 0.2 0 −0.2 −0.2 −0.2 0.2 0 −0.2 −0.2 0 −0.3 0
Acceptance_of_privacy_limits −0.2 −0.1 0.2 0.3 0.4 0.5 −0.2 0 0.5 0.3 0.2 0.3 0
CWA_download 0.3 −0.2 −0.3 0.2 0.1 0.3 0.5 −0.3 0 0.4 0.2 0 0.3 0.1
Figure A6: Pearson correlation matrix for variables in Figure 10, Wave 3. Positive correlations are displayed in red, negative correlations in blue, and small correlations in gray. Color intensity is proportional to the correlation coefficients. COVID-19 TRACKING: GERMANY 54
Correlation matrix: Wave 4
Gender_male
Government_announcements_use_and_trust −0.1
Social_media_use_and_trust 0 −0.1
Approval_of_German_pandemic_response −0.1 0.4 0
Belief_in_CWA_effectiveness 0.3 0.2 0.4 0
Age 0 0.1 −0.3 0 0.2
CWA_risk_of_harm_perception −0.1 −0.2 −0.3 0.1 −0.3 0 1.0
0.5 Trust_in_CWA_security −0.5 0.1 0.5 0.4 0 0.5 0.1 0.0
Find_government_guidelines_helpful 0.5 −0.3 0.1 0.4 0.5 0 0.5 0 −0.5
−1.0 COVID19_risk_perception 0.5 0.3 −0.2 0.1 0.3 0.3 0.1 0.4 −0.1
Attitudes_to_science_and_technology 0.2 0.2 0.4 −0.3 0.1 0.2 0.2 0 0.3 0.1
Belief_in_conspiracies −0.4 −0.4 −0.5 −0.6 0.4 −0.3 −0.3 −0.4 0.2 −0.5 −0.1
Libertarianism 0.2 0 −0.2 −0.2 −0.2 0.3 0 −0.1 −0.1 0 −0.2 0
Acceptance_of_privacy_limits −0.1 −0.2 0.1 0.3 0.4 0.5 −0.3 0 0.5 0.3 0.1 0.3 0
CWA_download 0.2 −0.1 −0.2 0.2 0.2 0.2 0.5 −0.3 0 0.4 0.2 0 0.2 0.1
Figure A7: Pearson correlation matrix for variables in Figure 10, Wave 4. Positive correlations are displayed in red, negative correlations in blue, and small correlations in gray. Color intensity is proportional to the correlation coefficients. COVID-19 TRACKING: GERMANY 55
Predictors of CWA download: Wave 3 Wave 4
Trust in CWA security
Perception of CWA effectiveness
Acceptance of privacy limits
COVID−19 risk perception
Belief in conspiracies
Social media use & trust
Government announcements use & trust
Approval of German pandemic response
Find government guidelines helpful
Attitudes to science and technology
Libertarianism
CWA risk of harm perception
Age
Gender: male (vs. female)
Education: low (vs. medium)
Education: high (vs. medium)
−1 0 1 2 3 0 1 2 Estimate Estimate Figure A8: Logistic regression models predicting Corona-Warn-App download in Waves 3 and 4 (an alternative representation of Figure 10). Dependent variable: downloaded the app (yes/no). Coefficients: measures from the survey (e.g., combined score for conspiracy beliefs; see Appendix Table B10). Horizontal point ranges indicate point estimates and 95% confidence intervals for each predictor. Education was dummy coded with the reference level “medium,” yielding two coefficients: low (vs. medium) and high (vs. medium) education. Following Gelman (2008), we standardized all continuous variables by two standard deviations (SD) and mean centered the binary gender variable. This way a 2-SD change in a continuous predictor variable is approximately equivalent to a change of category in a roughly balanced binary predictor variable (e.g., gender). In a logistic regression model, a slope reflects the relative change in log odds (while keeping all other predictors at their average values). Appendix Table A6 summarizes the regression results for these two models. Appendix Figures A6 and A7 display Pearson correlations for all variables in the regression model. COVID-19 TRACKING: GERMANY 56
Table A6 Regression Results for Figure 10 and Appendix Figure A8: Predictors of Corona-Warn-App (CWA) Downloads. Numbers represent point estimates and 95% confidence intervals for each predictor.
Dependent variable: CWA downloads Wave 3 Wave 4 (1) (2)
Trust in CWA security 2.386 2.148 (1.862,2.910) (1.657,2.638) Perception of CWA effectiveness 0.628 0.984 (0.247,1.009) (0.627,1.342) Acceptance of privacy limits 0.449 −0.104 (0.088,0.810) (-0.441,0.234) Approval of German pandemic response −0.342 0.039 (-0.769,0.085) (-0.322,0.400) Find government guidelenes helpful −0.039 −0.181 (-0.490,0.412) (-0.578,0.216) Social media use and trust 0.181 0.064 (-0.155,0.517) (-0.253,0.382) Government announcements use and trust 0.320 −0.124 (-0.099,0.739) (-0.517,0.270) Belief in conspiracies −0.122 0.083 (-0.501,0.257) (-0.371,0.538) Libertarianism −0.217 0.137 (-0.547,0.112) (-0.166,0.439) COVID-19 risk perception −0.001 0.398 (-0.355,0.352) (0.039,0.756) Attitudes to science and technology −0.405 −0.223 (-0.773,-0.038) (-0.560,0.114) Gender: male (vs. female) 0.354 0.194 (0.047,0.661) (-0.096,0.485) Age −0.155 −0.430 (-0.477,0.168) (-0.748,-0.111) CWA risk of harm perception −0.687 −0.403 (-1.088,-0.285) (-0.793,-0.013) Education: low (vs. medium) −0.710 −0.376 (-1.076,-0.344) (-0.723,-0.029) Education: high (vs. medium) −0.138 −0.066 (-0.565,0.289) (-0.474,0.343) Constant −0.575 −0.422 (-0.876,-0.275) (-0.707,-0.136)
Observations 1,183 1,140 COVID-19 TRACKING: GERMANY 57
Predictors of intention to download CWA: Wave 3 Wave 4
Trust in CWA security
Perception of CWA effectiveness
Acceptance of privacy limits
COVID−19 risk perception
Belief in conspiracies
Social media use & trust
Government announcements use & trust
Approval of German pandemic response
Find government guidelines helpful
Attitudes to science and technology
Libertarianism
CWA risk of harm perception
Age
Gender: male (vs. female)
Education: low (vs. medium)
Education: high (vs. medium)
−1 0 1 2 3 −1 0 1 2 3 Estimate Estimate Figure A9: Logistic regression models predicting intention to download the Corona-Warn-App in Waves 3 and 4 (an alternative representation of Figure 11). Dependent variable: downloaded the app (yes/no). Coefficients: measures from the survey (e.g., combined score for CWA in app security; combined score for conspiracy beliefs; see Appendix Table B10). Horizontal point ranges show point estimates and 95% confidence intervals for each predictor. Education was dummy coded with the reference level “medium,” yielding two coefficients: low (vs. medium) and high (vs. medium) education. Following Gelman (2008), we standardized all continuous variables by two standard deviations (SD) and mean centered the binary gender variable. This way a 2-SD change in a continuous predictor variable is approximately equivalent to a change of category in a roughly balanced binary predictor variable (e.g., gender). In a logistic regression model, a slope reflects the relative change in log odds (while keeping all other predictors at their average values). Appendix Table A6 summarizes the regression results for these two models. Appendix Figures A6 and A7 display Pearson correlations for all variables in the regression model. COVID-19 TRACKING: GERMANY 58
Table A7 Regression Results for Figure 11 and Appendix Figure A9: Predictors of Intention to Download the Corona-Warn-App (CWA). Numbers represent point estimates and 95% confidence intervals for each predictor.
Dependent variable: CWA downloads Wave 3 Wave 4 (1) (2)
Trust in CWA security 0.731 0.629 (0.090,1.372) (-0.092,1.350) Perception of CWA effectiveness 2.296 2.967 (1.735,2.856) (2.343,3.591) Acceptance of privacy limits 0.114 −0.203 (-0.410,0.637) (-0.777,0.371) Approval of German pandemic response 0.254 0.159 (-0.368,0.876) (-0.441,0.758) Find government guidelines helpful 0.028 0.113 (-0.636,0.691) (-0.549,0.774) Social media use and trust 0.406 0.315 (-0.074,0.886) (-0.190,0.819) Government announcements use and trust 0.737 0.222 (0.154,1.320) (-0.436,0.880) Belief in conspiracies −0.016 −0.670 (-0.576,0.544) (-1.449,0.109) Libertarianism −0.109 0.429 (-0.594,0.376) (-0.095,0.954) COVID-19 risk perception 0.252 0.475 (-0.265,0.769) (-0.171,1.120) Attitudes to science and technology −0.102 −0.114 (-0.628,0.424) (-0.677,0.449) Gender: male (vs. female) −0.171 0.039 (-0.613,0.270) (-0.436,0.515) Age −0.496 −1.008 (-0.947,-0.045) (-1.535,-0.482) CWA risk of harm perception −0.258 −0.475 (-0.806,0.291) (-1.132,0.183) Education: low (vs. medium) −0.351 0.393 (-0.881,0.180) (-0.195,0.980) Education: high (vs. medium) −0.333 0.539 (-0.998,0.332) (-0.168,1.245) Constant −1.526 −2.219 (-2.001,-1.050) (-2.782,-1.656)
Observations 758 684 COVID-19 TRACKING: GERMANY 59
Predictors of Trust in Corona−Warn−App (CWA) Security
Acceptance of privacy limits
COVID−19 risk perception
Belief in conspiracies
Social media use & trust
Government announcements use & trust
Approval of German pandemic response
Find government guidelines helpful
Wave 3
Attitudes to science and technology Wave 4
Libertarianism
CWA risk of harm perception
Gender: male (vs. female)
Age
Education: low (vs. medium)
Education: high (vs. medium)
−0.1 0.0 0.1 Estimate Figure A10: Linear regression models predicting trust in Corona-Warn-App security in Waves 3 and 4. Dependent variable: Trust in CWA security. Coefficients: various measures from the survey (e.g., combined score for conspiracy beliefs; see Appendix Table B10). Horizontal point ranges show point estimates and 95% confidence intervals for each predictor. Education was dummy coded with the reference level “medium” yielding two coefficients: low (vs. medium) and high (vs. medium) education. Following Gelman (2008), we standardized all continuous variables and the dependent variable by two standard deviations (SD) and mean centered binary variables. This way a 2-SD change in a continuous predictor variable is approximately equivalent to a change of category in a roughly balanced binary predictor variable (e.g., gender). Furthermore, because we also standardized the dependent variable by 2SD, a slope of, say, +0.1 can be interpreted as follows: If the predictor is increased by, for instance, 1SD of its distribution, the dependent variable increases by 0.1SD of its distribution (while keeping all other predictors at their average values). Appendix Table A8 summarizes the regression results for these two models. COVID-19 TRACKING: GERMANY 60
Table A8 Regression Results for Appendix Figure A10: Predictors of Trust in Corona-Warn-App Security in Waves 3 and 4. Numbers represent point estimates and 95% confidence intervals for each predictor.
Dependent variable: Trust in Corona-Warn-App security Wave 3 Wave 4 (1) (2)
Acceptance of privacy limits 0.279 0.248 (0.237,0.322) (0.205,0.292) Approval of German pandemic response 0.007 −0.003 (-0.044,0.058) (-0.050,0.044) Find government guidelines helpful 0.155 0.075 (0.100,0.211) (0.021,0.128) Social media use and trust 0.011 0.023 (-0.031,0.053) (-0.020,0.065) Government announcements use and trust 0.149 0.191 (0.097,0.200) (0.140,0.242) Belief in conspiracies −0.213 −0.195 (-0.259,-0.167) (-0.252,-0.139) Libertarianism 0.002 −0.044 (-0.039,0.043) (-0.085,-0.003) COVID-19 risk perception −0.027 −0.016 (-0.069,0.016) (-0.064,0.031) Attitudes to science and tech 0.154 0.115 (0.113,0.196) (0.073,0.158) CWA risk of harm perception −0.227 −0.265 (-0.271,-0.184) (-0.310,-0.220) Gender: male (vs. female) 0.006 0.080 (-0.033,0.045) (0.041,0.120) Age −0.014 0.019 (-0.055,0.027) (-0.024,0.061) Education: low (vs. medium) −0.015 −0.049 (-0.063,0.033) (-0.097,-0.001) Education: high (vs. medium) 0.003 0.040 (-0.054,0.061) (-0.017,0.097) Constant 0.007 0.017 (-0.032,0.047) (-0.022,0.056)
Observations 1,183 1,140 R2 0.575 0.572 Adjusted R2 0.570 0.566 COVID-19 TRACKING: GERMANY 61
Predictors of perceived effectiveness of Corona−Warn−App (CWA)
Acceptance of privacy limits
COVID19 risk perception
Belief in conspiracies
Social media use & trust
Government announcements use & trust
Approval of German pandemic response
Find government guidelines helpful Wave
Wave 3
Attitudes to science and technology Wave 4
Libertarianism
CWA risk of harm perception
Gender: male (vs. female)
Age
Education: low (vs. medium)
Education: high (vs. medium)
−0.1 0.0 0.1 Estimate Figure A11: Linear regression models predicting perceived effectiveness of the Corona-Warn-App in Waves 3 and 4. Dependent variable: Perceived effectiveness of the Corona-Warn-App. Coefficients: various measures from the survey (e.g., combined score for trust in CWA security; combined score for conspiracy beliefs; see Appendix Table B10). Education was dummy coded with the reference level “medium,” yielding two coefficients: low (vs. medium) and high (vs. medium) education. Following Gelman (2008), we standardized all continuous variables and the dependent variable by two standard deviations (SD) and mean centered binary variables. This way a 2-SD change in a continuous predictor variable is approximately equivalent to a change of category in a roughly balanced binary predictor variable (e.g., gender). Furthermore, because we also standardized the dependent variable by 2SD, a slope of, say, +0.1 can be interpreted as follows: If the predictor is increased by, for instance, 1SD of its distribution, the dependent variable increases by 0.1SD of its distribution (while keeping all other predictors at their average values). Appendix Table A9 summarizes the regression results for these models. COVID-19 TRACKING: GERMANY 62
Table A9 Regression Results for Appendix Figure A11: Predictors of Perceived Effectiveness of Corona-Warn-App in Waves 3 and 4. Numbers represent point estimates and 95% confidence intervals for each predictor.
Dependent variable: Perceived effectiveness of the Corona-Warn-App Wave 3 Wave 4 (1) (2)
Acceptance of privacy limits 0.316 0.286 (0.265,0.367) (0.233,0.340) Approval of German pandemic response 0.044 0.110 (-0.017,0.105) (0.052,0.168) Find governement guidelenes helpful 0.192 0.144 (0.126,0.259) (0.079,0.210) Social media use and trust 0.128 0.139 (0.079,0.178) (0.086,0.192) Governement announcements use and trust 0.089 0.114 (0.027,0.151) (0.051,0.177) Belief in conspiracies −0.034 0.015 (-0.089,0.021) (-0.055,0.085) Libertarianism 0.030 −0.027 (-0.020,0.079) (-0.078,0.023) Covid-19 risk perception 0.122 0.085 (0.071,0.173) (0.027,0.143) Attitudes to science and tech 0.076 0.078 (0.026,0.125) (0.025,0.130) CWA risk of harm perception 0.005 0.009 (-0.048,0.057) (-0.047,0.065) Gender: male (vs. female) 0.009 0.060 (-0.038,0.056) (0.012,0.109) Age −0.077 −0.055 (-0.126,-0.027) (-0.108,-0.002) Education: low (vs. medium) 0.003 −0.063 (-0.054,0.060) (-0.122,-0.003) Education: high (vs. medium) −0.003 −0.017 (-0.072,0.065) (-0.087,0.054) Constant −0.001 0.037 (-0.048,0.046) (-0.012,0.085)
Observations 1,183 1,140 R2 0.394 0.344 Adjusted R2 0.386 0.336 COVID-19 TRACKING: GERMANY 63
Table A10 Demographic Information: Education and Region
Wave 1 Wave 2 Wave 3 Wave 4
Sample size N 829 1,109 1,231 1,188 Education (highest level) (%) University 25.8 23.2 21.5 21.7 Abitur (academic track) 26.8 27.8 23.5 26.1 Realschule (intermediate track) 33.3 35.3 36.6 36.1 Hauptschule (vocational track) 13.5 13.1 17.6 15.5 None 0.6 0.7 0.7 0.6 Region (%) Bremen, Hamburg, Lower Saxony, Schleswig-Holstein 16.2 16.3 16.5 16.6 North Rhine-Westphalia 22.7 22.2 21.3 23.1 Hesse, Rhineland-Palatinate, Saarland 13.9 15.6 14.5 13.3 Baden-Württemberg 11.2 9.8 12.0 10.4 Bavaria 14.8 14.6 14.6 14.8 Berlin, Brandenburg, Mecklenburg-Vorpommern, Saxony-Anhalt 13.8 14.2 13.4 13.8 Saxony, Thuringia 7.5 7.3 7.7 8.0 COVID-19 TRACKING: GERMANY 64
Appendix B Supplementary Information: Items and Covariates
Table B1 Items Assessing Impact of COVID-19, and Information Use and Trust
Question Answer options
Impact of COVID-19 Have you tested positive for COVID-19? Yes/No Do you know anyone who has tested positive for COVID-19? Yes/No How many days, if any, did you voluntarily go into quarantine? Open numeric response on slider scale Have you temporarily or permanently lost your job as a consequence of the COVID-19 pan- Yes/No demic? Information Use and Trust How often do you rely on the following media to stay informed about developments surround- 5-Point Likert Scale (1 = Never, 5 = ing the COVID-19 pandemic? Please select all media you use: Newspapers (also online); Always) Social media; Friends and family; Radio; Television; Government announcements (including their websites) How would you rate the following sources? Do you think the information about the COVID- 5-Point Likert Scale (1 = Not at All, 5 19 pandemic is accurate and trustworthy? Newspapers (also online); Social media; Friends = Completely) and family; Radio; Television; Government announcements (including their websites) COVID-19 TRACKING: GERMANY 65
Table B2 Items Assessing Perceived Risks of COVID-19 on a 5-Point Likert Scale (1 = Not at All, 5 = Extremely)
Question Label
How severe do you think the novel coronavirus (COVID-19) will be for the general population? Severity for German population How harmful would it be for your health if you were to become infected COVID-19? Harm to your health if infected How concerned are you that you might become infected with COVID-19? Concern you will be infected How concerned are you that somebody you know might become infected with COVID-19? Concern someone you know will be infected
Table B3 Items Assessing Government Response, Fatalities, and Compliance with Social Distancing Policies
Question Answer options
Government response (only Waves 3 and 4) How understandable do you find the government’s current policies and actions related to the 5-Point Likert Scale (1 = Not at All, 5 COVID-19 pandemic? = Completely) How helpful do you find the government’s current approach to the COVID-19 pandemic? 5-Point Likert Scale (1 = Not at All, 5 = Extremely) Overall, how well do you think the governments of the following countries have handled the COVID-19 pandemic so far? (Wave 3: Australia, China, Germany, Italy, Singapore, South Korea, Spain, United States, 5-Point Likert Scale (1 = Extremely, 5 United Kingdom) = Not at All) (Wave 4: Australia, China, Germany, Italy, Singapore, South Korea, Spain, United States, 5-Point Likert Scale (1 = Extremely, 5 United Kingdom, Brazil, Russia, India, Sweden, France) = Not at All) Fatalities (only Waves 2 and 3) Please estimate the number of deaths caused by COVID-19 in each of the following countries Slider for numeric entry. (Australia, China, Germany, Italy, Singapore, South Korea, Spain, United States, United Kingdom) Compliance (Wave 2 and 3) In your opinion, what percentage of the population complies with the government’s policy Open numeric entry. on social distancing? How much do you follow the government’s policy on social distancing? Multiple choice: I do not follow these guidelines at all; I follow these guide- lines a little; I follow these guidelines somewhat; I follow these guidelines mostly, but not completely; I follow these guidelines completely; I go a lit- tle beyond what is required by govern- ment policy; I go a little beyond what government policy dictate; I go well be- yond what government policy dictates; I am under complete quarantine and never leave home Compliance (AHA+L, Wave 4) The federal government recommends the following combination of preventive measures to Yes/No/Partially contain the COVID-19 pandemic, known as AHA + L rules: Keep your distance, observe hygiene, wear an everyday mask, and ventilate regularly. Have you heard about the AHA+L rules? How much do you follow these AHA+L rules? (Keep distance (Keep at least 1.5 meters away 5-Point Likert Scale (1 = Not at All, 5 from fellow humans); Observe hygiene (Cough and sneeze properly and wash or disinfect = Completely) hands regularly); Wear everyday mask (If crowded: wear mouth/nose protection); Airing closed rooms regularly.) What percentage of the German population do you think adheres fairly or completely to the Open numeric entry. AHA+L (distance, hygiene, everyday mask + ventilation) rules? Please estimate. COVID-19 TRACKING: GERMANY 66
Table B4 Scenarios Used in the Study (Translated from the German)
Scenario Description Wave
Severe The COVID-19 pandemic has rapidly become a worldwide threat. Many experts agree that slowing 1, 2 the spread of the virus is essential to minimize the impact on the health care system and the economy, and to save many lives. The government might consider using mobile phone data to identify and contact those who may have come into contact with people with COVID-19. All people using a mobile phone would be included in the project, with no possibility of opting out. Data would be stored in an encrypted format on a secure server accessible only to the government, which may use the data to locate people who violate lockdown orders and fine or arrest them where necessary. Data would also be used to help shape the public health response and to contact people who might have been exposed to COVID-19. Individual quarantine orders could be made on the basis of this data.
Mild The COVID-19 pandemic has rapidly become a worldwide threat. Many experts agree that slowing 1,2 the spread of the virus is essential to minimize the impact on the health care system and the economy, and to save many lives. The government could consider using mobile phone data to identify and contact those who may have come into contact with people with COVID-19. Only people who download a government app and agree to be tracked and contacted would be included in the project. The more people who download and use this app, the more effectively the government would be able to contain the spread of COVID-19. Data would be stored in an encrypted format on a secure server accessible only to the government. Data would only be used to contact those who might have been exposed to COVID-19.
Bluetooth The COVID-19 pandemic has rapidly become a worldwide threat. Many experts agree that slowing 2 the spread of the virus is essential to minimize the impact on the health care system and the economy, and to save many lives. Apple and Google have proposed adding a contact tracing capability to existing smartphones to inform people who have been exposed to others with COVID-19. This would help reduce community spread of COVID-19 by enabling people to voluntarily self-isolate. When two people are near each other, their phones would connect via Bluetooth. If a person is later identified as being infected, the people to whom they have been in close proximity are then notified without the government knowing who they are. The use of this contact tracing capability would be completely voluntary. People who are notified would not know the identity of the person who had tested positive.
Corona- The Corona-Warn-App app is designed to help detect and break infection chains at an early stage. 3,4 Warn-App Currently, local health authorities are trying to trace infection chains. With the app, this process can be automated and thus unfold much faster and more accurately. Users can be warned immediately if they have been in the vicinity of an infected person. The app was developed by Deutsche Telekom and SAP and published by the Robert Koch Institute. It records which smartphones have come in proximity to each other. To do this, smartphones with the app exchange randomly generated encryption keys via Bluetooth. The distance is estimated on the basis of the signal strength. If a user tests positive for COVID-19, they can share their test result in the app in order to inform users who have been in their vicinity. Infected users are explicitly asked whether they want to share their result for contact tracing. As an alternative to digital transmission, validation is available via a call center. Every 24 hours, the app checks whether the user has had contact with a person who has registered an infection on the app. The app does not evaluate any geodata and does not transmit any location information. The developers also assure that no personal data is sent or stored. The anonymized contact data is not stored centrally, but locally on the user’s smartphone. The comparison of whether an infected person has been encountered is carried out locally on the smartphone. No data leaves the phone for matching, according to the developers. Only the anonymized list is stored centrally and regularly retrieved by the smartphones to identify possible infectious encounters. COVID-19 TRACKING: GERMANY 67
Table B5 Items Assessing Acceptability of Tracking in the Three Hypothetical Scenarios and Downloads/Use of the Corona-Warn-App (CWA)
Scenario Question Label Answer options
Severe Is the use of cell phone data for location tracking acceptable in this scenario? Acceptability of Yes/No the scenario Severe Would your decision change if the government was required to delete the data With follow up: Yes/No and stop tracking after 6 months? delete data Severe Would your final decision change if there was an option to opt out of data With follow up: Yes/No collection? opt out Mild If, as depicted in this scenario, the government developed a tracking app to Acceptability of Yes/No help reduce the spread of COVID-19, would you download and use it? the scenario Mild Would your decision change if the government was required to delete the data With follow up: Yes/No and stop tracking after 6 months? delete data Mild Would your final decision change if data was only stored on your smartphone With follow up: lo- Yes/No (not on government servers) and you had the option to make this data avail- cal data able if you tested positive for COVID-19? Bluetooth If, as depicted in this scenario, Apple and Google added a COVID-19 contact Acceptability of Yes/No tracing capability to smartphones, would you use it? the scenario Bluetooth Would your decision change if Apple and Google promised to delete all data With follow up: Yes/No and remove the contact tracing system after 6 months? delete data CWA Have you downloaded the Corona-Warn-App? CWA downloads Yes/No CWA Will you download the Corona-Warn-App in the future? With follow up: fu- Yes/No ture downloads CWA Is the Corona-Warn-App still installed on your phone? – Yes/No CWA Do you generally have Bluetooth switched on so the Corona-Warn-App can – Yes/No/Only when I leave operate effectively? the house/I don’t know CWA Have you made any attempts to convince your friends and/or family to down- – Yes/No load the Corona-Warn-App? CWA Will you download the Corona-Warn-App in the future? – Yes/No CWA Do you think that the government should make the Corona-Warn-App – Yes/No mandatory? COVID-19 TRACKING: GERMANY 68
Table B6 Items Assessing the Perceived Effectiveness and Risks of Hypothetical Tracking Apps in Different Scenarios (Waves 1 and 2) and of the Corona-Warn-App (CWA, Waves 3 and 4) on a 6-Point Likert Scale (1 = Not at All, 6 = Very).
Question Label Scenario
Waves 1 and 2 How confident are you that the government app would reduce your likelihood of con- Reduce Likelihood to Severe, Mild, tracting COVID-19? Contract Bluetooth How confident are you that the government app would reduce the spread of COVID-19? Reduce Spread Severe, Mild, Bluetooth How confident are you that the government app would help you resume your normal Return to Activity Severe, Mild, activities more rapidly? Bluetooth How easy is it for people to decline participation in the proposed project? Ease of Declining Partic- Severe, Mild, ipation Bluetooth How serious is the risk of harm that could arise from the proposed project? Risk of Harm Severe, Mild, Bluetooth How much do you agree that the government [Apple and Google for Bluetooth scenario] Only Necessary Data Severe, Mild, is collecting only necessary data? Collection Bluetooth How sensitive is the data being collected in the proposed project? How Sensitive Are Data Severe, Mild, Bluetooth How much do you trust the government to use the tracking data only to deal with the Trust: Data for Pan- Severe, Mild, COVID-19 pandemic? demic Only Bluetooth How much do you trust the government to be able to ensure the privacy of each Trust: Privacy Protec- Severe, Mild, individual? tion Bluetooth How secure is the data that would be collected for the proposed project from access Trust: Security From Severe, Mild, by third parties ? 3rd Party Bluetooth To what extent do people have ongoing control of their data? User Control Over Data Severe, Mild, Bluetooth How confident are you that other citizens like yourself would be able to download and Others’ Ability to Down- Mild, Blue- effectively use the app? load the App tooth Waves 3 and 4 How likely do you think it is that the Corona-Warn-App will reduce your risk of coming Reduce Likelihood to CWA in contact with COVID-19? Contract Has the Corona-Warn-App already helped you to resume your normal activities? Return to Activity Past CWA How confident are you that the Corona-Warn-App will help you to maintain your Return to Activity Fu- CWA normal activities in the future course of the pandemic? ture To what extent do you think the Corona-Warn-App has already reduced the spread of Reduce Spread Past CWA COVID-19? How confident are you that the Corona-Warn-App will reduce the spread of COVID-19? Reduce Spread Future CWA How easy is it for people to decline participation in the Corona-Warn-App? Ease of Declining Partic- CWA ipation How serious is the risk of harm that could arise from the proposed project? Risk of Harm CWA How much do you trust that the Corona-Warn-App is collecting only necessary data? Only Necessary Data CWA Collection How sensitive is the data being collected by the Corona-Warn-App? How Sensitive Are Data CWA How much do you trust that the data will only be used to deal with the COVID-19 Trust: Data for Pan- pandemic? demic Only How much do you trust the app to be able to ensure the privacy of each individual? Trust: Privacy Protec- CWA tion How secure is the data collected by the Corona-Warn-App from access by third parties? Trust:Security From 3rd CWA Party To what extent do people have ongoing control of their data? This includes control User Control Over Data CWA over how and when the data is collected, as well as access to view and delete the data after collection. How confident are you that other citizens like yourself are able to download and effec- Others’ Ability to Down- CWA tively use the Corona-Warn-App? load the App COVID-19 TRACKING: GERMANY 69
Table B7 Reasons (Not) to Download the Corona-Warn-App (Multiple Responses Allowed)
Response Option Label
Why did you download the Corona-Warn-App? Please select all that apply. To return to normal activities To return to normal activities To protect my health To protect my health To protect the health of others To protect the health of others To follow government recommendations To follow government recommendations To help the economy To help the economy Other reasons Why won’t you download the Corona-Warn-App? Please select all that apply. I am concerned about privacy. Privacy concerns I don’t trust the government. Lack of gov trust I am worried about battery usage on my phone. Concerns: Battery usage I don’t think it will be effective. Belief it is not effective I am worried about normalizing government tracking. Concerns: Normalizing gov tracking I am concerned about civil liberties. Concerns: Civil liberties I don’t own a smartphone. Don’t own a smartphone My phone is too old to run the app. Phone too old I am concerned about others gaining access to my data. Concerns: 3rd party access I think that the virus doe not present a particular danger (Only Wave 4) Belief virus is not dangerous Other reasons (Only Wave 4) Other reasons
Table B8 Items Assessing Worldviews
Name Question Scale
Free market 1. An economic system based on free markets unrestrained by government interfer- 7-point Likert scale: 1 attitudes (Liber- ence automatically works best to meet human needs. 2. The free market system may = strongly disagree, 7 = tarianism) be efficient for resource allocation but it is limited in its capacity to promote social strongly agree justice. (reverse coded) 3. The government should interfere with the lives of citizens as little as possible.
Attitudes to sci- 1. Science and technology are making our lives healthier, easier, and more comfort- 7-point Likert scale: 1 ence and tech- able. 2. Because of science and technology, there will be more opportunities for the = strongly disagree, 7 = nology next generation. strongly agree
Belief in conspir- 1. There are secret organizations that have great influence on political decisions. 7-point Likert scale: 1 acies (Wave 3) 2. Most people do not see how much of our lives are determined by plots hatched = strongly disagree, 7 = in secret. 3. There are certain political circles with secret agendas that are very strongly agree . influential. 4. I think that the various conspiracy theories circulating in the media are absolute nonsense. (reverse coded) 5. Secret organizations can manipulate people psychologically so that they do not notice how their life is being controlled by others. 6. Selfish interests have conspired to convince the public that COVID-19 is a major threat.
Beliefs in con- 1. COVID-19 does not really exist. It is a myth created by some influential people 5-point Likert scale: 1 = spiracies (Wave or institutions. 2. COVID-19 was created in a laboratory and deliberately released undoubtedly false, 5 = un- 4) to achieve geopolitical or economic goals. 3. There is a link between 5G and the doubtedly true spread of COVID-19. 4. The severity of COVID-19 is overstated. The actual risk is not higher than that of a seasonal influenza. 5. The government exaggerates the seriousness of the pandemic in order to divert attention from other problems within Germany. 6. Wearing masks (mouth and nose protection) protects oneself and others from contracting a COVID-19 infection. (reverse coded) 7. COVID-19 was created by pharmaceutical companies to benefit from the need for a vaccine. COVID-19 TRACKING: GERMANY 70
Table B9 Items Assessing the General Acceptability of Privacy-Encroaching Measures During the Pandemic on a 4-Point Likert Scale (1 = Very Acceptable, 4 = Not Acceptable at All). Question: How Acceptable Is it For the Government to Take the Following Measures to Limit the Spread of the Virus During the COVID-19 Pandemic?
Item Label
Provide access to the medical records of individuals. Access to medical records Track people’s locations using their smartphone data. Location tracking data Enable temporary relaxation of data protection regulations. Relaxation of data protection regulations Collect data about personal contacts and interactions. Data on contacts & interactions Enforce people to use an app that notifies when those in quarantine leave the house. Notification when people leave quarantine
Collect data on the infection and immunity status of citizens. Data on infections & immunity status COVID-19 TRACKING: GERMANY 71
Table B10 Variables and Items for Regression models
Variable Items Scale
CWA download Have you downloaded the Corona-Warn-App? Yes (1), No (0)
Trust in CWA secu- 1. How much do you trust the government to use the Corona-Warn-App data 6-point Likert scale: 1 = not at rity only to deal with the COVID-19 pandemic? 2. How much do you trust that all, 6 = very the Corona-Warn-App can ensure the privacy of each individual who uses it? 3. How much do you trust that the Corona-Warn-App is collecting only necessary data? 4. How secure is the data collected by the Corona-Warn- App?
Perception of CWA 1. How confident are you that the Corona-Warn-App will help you to maintain 6-point Likert scale: 1 = not at effectiveness your normal activities in the future course of the pandemic? 2. How confident all, 6 = very are you that the Corona-Warn-App will reduce the spread of COVID-19? 3. How likely do you think it is that the Corona-Warn-App will reduce your risk of coming into contact with COVID-19?
CWA risk of harm How serious is the risk of harm that could arise from the Corona-Warn-App? 6-point Likert scale: 1 = not at perception all, 6 = very
Acceptance of pri- How acceptable is it for the government to take the following measures to 4-point Likert scale: 1 = not vacy limits limit the spread of the virus during the COVID-19 pandemic? Combined acceptable at all, 4 = very ac- measure for 6 items: See Appendix Table B9 ceptable
Social media use & 1. How often do you rely on the following media (social media) to keep you 5-point Likert scale: 1 = trust informed about the developments surrounding the COVID-19 pandemic? 2. never/not at all, 5 = al- How do you rate the following sources (social media)? 3. Do you think that ways/completely the information on the COVID-19 pandemic is correct and trustworthy?
Government an- 1.How often do you rely on the following media (government announcements) 5-point Likert scale: 1 = nouncements use & to keep you informed about developments surrounding the COVID-19 pan- never/not at all, 5 = al- trust demic? 2. How do you rate the following sources(government announce- ways/completely ments)? Do you think that the information on the COVID-19 pandemic is correct and trustworthy?
COVID-19 risk per- Combined measure for 4 items: See Appendix Table B2 5-point Likert scale: 1 = not at ception all, 5 = extremely
Approval of Ger- How well overall do you think the governments in the following countries have 5-point Likert scale: 1 = not at man pandemic re- handled the COVID-19 pandemic? - Germany all, 5 = extremely sponse
Find government How helpful are the government guidelines in deciding how to act in relation 5-point Likert scale: 1 = not at guidelines helpful to the COVID-19 pandemic? all, 5 = extremely
Attitudes to science Combined measure for 2 items: See Appendix Table B8 7-point Likert scale: 1 = and technology strongly disagree, 7 = strongly agree
Free market atti- Combined measure for 3 items: See Appendix Table B8 7-point Likert scale: 1 = tudes (libertarian- strongly disagree, 7 = strongly ism) agree
Belief in conspira- Combined measure for 6 items: See Appendix Table B8. 7-point Likert scale: 1 = cies (Wave 3) strongly disagree, 7 = strongly agree
Belief in conspira- Combined measure for 7 items: See Appendix Table B8 5-point Likert scale: 1 = un- cies (Wave 4) doubtedly false, 5 = undoubt- edly true
Gender (male) What gender do you identify with? (Note: "Other" category was excluded Male (1), Female (0) from regression analyses due to its small sample size, see Table1)
Education Please indicate your highest level of education. “Low” = None, Hauptschule, Realschule; “Medium” = Abitur, “High” = University
Age How old are you? Free numeric text entry