Path to Citizenship: Data-driven versus User-driven Outreach Campaigns to Boost Naturalization Rates

This RCT builds on the results of the first experiment (AEA RCT Registry, trial ID #0015790), which tested four different informational interventions in the form of leaflets. The leaflets addressed four perceived hurdles based on insights from a previously evaluated campaign, focus group discussions with different immigrant groups, and expert interviews. These included the perceived complexity regarding the time and paperwork needed to apply, the unawareness of available support options, the lack of knowledge about application requirements, and the perceived lack of benefits.

This second experiment tests two different approaches to delivering the same information to citizenship-eligible immigrants. The first approach is data-driven, relying on machine learning to identify the most effective leaflet for each recipient. We used the data from the first experiment to estimate the conditional average treatment effects of the four leaflets and then predict the expected treatment effect associated with each leaflet for all individuals in the treatment group, based on their observable characteristics. We then used a plug-in rule to assign to each individual the leaflet with the highest predicted treatment effect.

The second approach is user-driven. We have developed a new leaflet that includes brief summaries of the four leaflets used in the initial experiment, along with personalized QR codes and short URLs that link to downloadable PDF versions of each leaflet. The short summaries were designed to highlight which questions the recipient could find answers to in each leaflet, guiding their information acquisition. We designed two versions of this leaflet, altering the order in which the four leaflets were introduced to test for order effects. QR-codes and short URLs are trackable at the treatment group level, with the group identified by gender, country of origin (German-speaking versus non-German-speaking), and order version. For each of the four original leaflets, we will be able to track the number of clicks at the treatment group level.

The leaflets are sent to citizenship-eligible immigrants in the treatment groups by the City of Zurich, along with a letter from the mayor informing them that they meet the eligibility criteria and encouraging them to apply. The letter from the city mayor addresses the perceived unwelcomeness to apply for naturalization, which can act as a barrier to initiating the process. The four leaflets, instead, target specific informational and psychological barriers.

Preliminary results from the first experiment show that all four leaflets significantly increased citizenship applications approximately six months later, with an average treatment effect of 3.2 percentage points from a control mean of 7%. The two interventions in the second RCT will allow us to contrast the effectiveness of a flexible data-driven approach available to policy makers, and based solely on immigrant background characteristics readily available from register data, with an approach that does not directly deliver information to citizenship-eligible immigrants but encourages them to acquire the information based on their perceived hurdles and needs.

2025-07-08

2026-01-09

The main outcome of interest is whether a participant applies for naturalization based on administrative records, measured approximately 6 months after the intervention.

The administrative data used to measure the primary outcome will be shared with the research team only after the intervention window ends in January 2026.

Naturalization applications: The City of Zurich and the Canton of Zurich provide us with administrative records of whether and when study participants submit an application for Swiss citizenship. This is a dummy variable that takes the value of one if the participant has submitted a naturalization application by the primary time point.

Our primary time point of interest is approximately 6 months after the intervention. After this time point, an intervention will be selected based on the results of both experiments to be used for participants in the pure control group to receive information.

We will also study the impact of the two information treatments on individual beliefs and integration outcomes, as measured through a two-wave survey designed to complement the analysis of administrative records. The endline survey (second wave) was disseminated approximately 6 weeks after the intervention (9 months after the first experiment’s interventions). This will allow testing for short-term effects on self-reported measures of integration, marginalization, and the expected costs and benefits of naturalization. We will focus on the following set of secondary outcomes based on survey questions:

Submitted Application: a binary indicator taking value one if the participant responded Yes to the survey question asking whether they submitted an application.
Naturalization Decision Index.
Naturalization Complexity Index.
Naturalization Support Index.
Naturalization Knowledge Index.
Naturalization Benefits Index.
Inclusion Index.
Psychological Integration Index.
Social Integration Index.
Political Integration Index.
Economic Intregration Index
Linguistic Integration Index
Other perceived obstacles.

All indices are obtained by extracting the first principal component using polychoric principal component analysis of the corresponding sets of survey items detailed in the pre-analysis plan.

Sample Selection and Recruitment

The target group for the second experiment consists of all the citizenship-eligible immigrants who were randomly assigned to the pure control group in the first experiment and did not submit a naturalization application by mid-April 2025. The initial sample included the universe of individuals who became eligible for naturalization between July 2021 and December 2024 and did not submit a naturalization application nor were part of the previous outreach campaign. This group was identified with the support of the City of Zurich based on the following naturalization criteria: individuals who hold a type-C residence permit, have lived in Switzerland for at least 10 years and in Zurich for at least two years, and are aged 16 or older.

The survey invitation (for both the baseline and endline surveys) was also mailed to all newly naturalization-eligible immigrants in the City of Zurich. The invitation letters and survey are available in six languages (English, French, German, Italian, Portuguese, and Spanish) to facilitate participation in the survey. Individuals receive the invite in German and a second language matched to their nationality. The survey can be completed online or in a paper-and-pencil format to facilitate access for participants with varying levels of digital literacy. Importantly, the survey was sent out by the research team with a time lag relative to the interventions to make it ostensibly independent from the outreach campaigns and minimize experimenter demand effects. Lastly, to improve survey participation, two reminders were also sent out.

Experimental Design

The study employs a staggered design, with the experimental timeline divided into two phases. In a preliminary step, the City of Zurich identified the newly citizenship-eligible residents that the outreach campaign will target. During Phase 2, which corresponds to the second RCT, participants are randomly allocated to the following treatment arms:
Data-driven treatment;
User-driven treatment;
Pure control group.

Participants in the data-driven treatment group receive one of the four information leaflets used in Phase 1, according to a plug-in rule that assigns to everyone the leaflet with the highest predicted treatment effect on naturalization based on their observed characteristics available in register data. Participants in the user-driven treatment receive a newly designed leaflet that briefly describes the four leaflets used in Phase 1, including four personalized QR codes and short URLs that link to downloadable PDF versions of each leaflet.

The original Phase 1 leaflets address four perceived hurdles based on insights from a previously evaluated campaign, focus group discussions with various immigrant groups, and expert interviews. These include the perceived complexity regarding the time and paperwork needed to apply, the unawareness of available support options, the lack of knowledge about application requirements, and the perceived lack of benefits.

Participants in the two treatment groups receive the leaflets by mail, along with a letter from the City mayor informing them that they meet the eligibility criteria and encouraging them to apply. The letter addresses the perceived unwelcomeness of applying for naturalization, which can serve as an additional barrier to starting the naturalization process. The City of Zurich sent the information letters with the two randomly assigned interventions to the recipients in the treatment groups during the second week of July 2025.

Participants in the pure control group will not receive any informational materials during Phase 2 and will receive information after the results from Phase 2 are available, in January 2026. The information delivery approach that will be used for the pure control group will be informed by the results of both experiments. The Phase 2 treatments aim to compare the effectiveness of a prescriptive and targeted approach to information provision to an untargeted information acquisition approach driven by participants.

The unit of randomization is the street-address cluster. Cluster-level randomization ensures that everyone living in the same apartment building will receive the same treatment, minimizing the scope for unintended spillovers across treatment groups or between treatment and control groups. In addition, the sample is stratified according to five regions of origin, which group together similar countries of origin. These regions could influence the primary outcome – naturalization applications – by affecting the perceived hurdles and incentives to apply for citizenship. Stratified randomization ensures a balanced distribution of regions of origin across experimental groups, reducing the potential for confounding by chance.

To complement the analysis of register data on applications for Swiss citizenship with data about recipients’ beliefs, we designed a two-wave survey. The first-wave baseline survey was disseminated to all study participants in August 2024. The second-wave endline survey was disseminated to all study participants in August 2025, approximately 9 months (6 weeks) after treated participants in the Phase 1 (Phase 2) RCT received the information treatments.

Outcome Data Collection

We will collect administrative data and survey responses. The primary outcome of interest is whether a participant applies for naturalization, based on administrative records provided by the Canton of Zurich, on whether and when participants apply for Swiss citizenship. To estimate the effects of Phase 2 interventions, we will analyze application data approximately six months after letters were sent out to participants in the four treatment groups (mid-December 2025 – beginning of January 2026).

The timing of the endline survey will allow us to complement the analysis of application data with data on individuals’ beliefs 1.5 months after Phase 2 interventions. The survey covers questions on socio-demographics, integration, marginalization, and the expected costs and benefits of naturalization. Using items from the IPL-12 integration index (Harder et al., 2018), we will measure immigrants’ economic, linguistic, political, psychological, social, and navigational integration. These short modules (two to four questions each) are designed to capture key dimensions of integration, understood as immigrants’ capacity and knowledge to build a self-determined and successful life.

Randomization done by statistical software

The unit of randomization is the street-address cluster. Cluster-level randomization ensures that everyone living in the same apartment building will receive the same treatment, minimizing the scope for unintended spillovers across treatment groups or between treatment and control groups.

Yes

Total number of clusters in the experimental sample: n = 2582

The study population includes N=3878 citizenship-eligible individuals identified by the City of Zurich. This comprises the set of individuals assigned to the pure control group in the first experiment, who did not submit a citizenship application by mid-April 2025.

During Phase 2, participants are randomly allocated to two treatment arms or a pure control group according to the following schedule:

Group B: data-driven information treatment (33.3% of the sample, N=1328, clusters=862)
Group C: user-driven information treatment (33.3% of the sample, N=1269, clusters=860)
Group D: pure control group (33.3% of the sample, N=1281, clusters=860)