Preferences and a Policy Experiment on Regional Job Choices

How to cite

Choi, Syngjoo; Kwon, Ohik; Kim, Seonghoon; Lee, Seunghak. 2025. "Preferences and a Policy Experiment on Regional Job Choices." AEA RCT Registry. AEARCTR-0016482. https://doi.org/10.1257/rct.16482-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Syngjoo Choi* , Ohik Kwon , Seonghoon Kim , Seunghak Lee

* Primary investigator

Registry ID
AEARCTR-0016482
DOI
https://doi.org/10.1257/rct.16482-1.0
First registered
Last updated
Status
On going
Trial dates
2025-08-01 – 2027-12-31
Countries
Private
JEL codes
C93, J28, J61, R23, R58
Registry record
http://www.socialscienceregistry.org/trials/16482

Abstract

South Korea faces significant regional inequality, with young adult workers increasingly concentrated in the Seoul metropolitan area despite government efforts to promote balanced regional development. This study investigates the factors influencing young adult workers' (aged 19-39) decisions to work in non-Seoul urban areas, addressing a critical challenge for regional sustainability and economic vitalization. Using a sample of 3,200 young wage workers stratified by age, gender, and region, we conduct two complementary experiments designed to inform evidence-based policy interventions.
First, we implement a discrete choice experiment (DCE) where participants make 16 binary choices between hypothetical job offers (8 in the general module and 8 in the Pohang-specific module). The experiment systematically varies workplace attributes across key dimensions: distance from Seoul (travel time), housing support policies, work flexibility (worktime and workplace flexibility), workplace culture, and regional amenities attributes (education, medical, commercial, transport, and leisure), along with annual income. Income levels are randomly drawn from 70-150% of each participant's current or expected salary to ensure realistic trade-offs. The Pohang-specific module features 8 choices with fixed locations (Seoul metropolitan area versus Pohang) and an additional industry sector attribute, enabling place-specific policy insights.
Second, we conduct a randomized information treatment experiment to test whether providing information about regional living conditions can shift young adult employment preferences. Participants are randomly assigned to view one of three 30-40 seconds long videos: factual housing cost comparisons between Seoul and regional cities (Treatment 1), an AI-generated conversational dialogue discussing regional quality of life and cost advantages (Treatment 2), or a placebo video about tourism (Control). We then measure changes in intentions to work in non-Seoul region areas, sector-specific willingness to work for regional public institutions and large corporations, and corrections in misperceptions about regional housing costs.
By incorporating specific policy attributes in the choice experiment, including various forms of housing support and work arrangements, this study enables pre-implementation evaluation of policy effectiveness. The experimental design allows us to calculate willingness-to-pay estimates for each attribute, revealing the implicit price young workers place on different job and location characteristics. Furthermore, our stratified sampling approach enables analysis of heterogeneous treatment effects across key demographic dimensions including current residence (metropolitan versus regional), prior regional experience, age cohorts, and employment status. This heterogeneity analysis is crucial for identifying which population segments are most responsive to different policy interventions, enabling more targeted and cost-effective policy design.
The study contributes to both academic literature and policy practice by providing causally identified estimates of young adult preferences for regional employment and testing scalable information interventions that could shift these preferences. By linking stated preferences to specific policy instruments, we generate actionable evidence for policymakers seeking to attract young talent to regional areas and promote more balanced spatial development in South Korea.

Intervention

This study implements two distinct experimental interventions to understand young adult employment preferences for regional areas in South Korea.

The first intervention is a discrete choice experiment where participants complete 16 binary choice tasks between hypothetical job offers (8 in the general module and 8 in the Pohang-specific module). Each choice varies systematically across eleven attributes: Seoul accessibility measured by travel time, regional housing support policies including subsidies and settlement funds, work time flexibility options, workplace flexibility, workplace culture from hierarchical to horizontal, local education infrastructure quality, medical facility accessibility, commercial amenity availability, public transportation convenience, cultural and leisure facilities, and annual income. Income levels are randomized between 70-150% of each participant's current or expected salary. Additionally, participants complete a specialized Pohang module with 8 choices comparing fixed locations of Seoul metropolitan area versus Pohang, incorporating an additional industry sector attribute.

The second intervention consists of an information treatment experiment using randomized video assignments. Participants view one of three videos lasting 30-40 seconds: Treatment 1 presents factual housing cost comparisons between Seoul and regional cities, Treatment 2 features AI-generated dialogue discussing regional quality of life and cost advantages, and the Control group views a placebo video about tourism destinations. Following video exposure, we measure changes in regional work intentions and housing cost perceptions.

Experimental design

Our study employs two complementary experimental designs with a sample of 3,200 young wage workers aged 19-39, stratified by age, gender, and region.

[Discrete Choice Experiment]
Participants complete 16 binary choices between pairs of hypothetical job offers (8 in the general module and 8 in the Pohang-specific module). We systematically vary eleven attributes across choices using an efficient design that ensures orthogonality and level balance. The income attribute is personalized for each participant, randomly drawn from 70-150% of their current or expected salary to ensure realistic trade-offs. The Pohang-specific module with 8 choices features fixed location comparisons and an industry sector attribute.

[Information Treatment Experiment]
Following the DCE, participants are randomly assigned with equal probability to one of three video conditions. Random assignment occurs at the individual level through the survey platform's randomization algorithm. We measure treatment effects on regional work intentions and housing cost perceptions immediately after video exposure.

[Pilot Study]
A pilot study with 204 participants was conducted in August 2025. Based on pilot results, we added a workplace flexibility attribute to the general DCE module and expanded the Pohang-specific module from 4 to 8 choices to enable more precise estimation of location-specific preferences. All other instruments and procedures were validated for the main study.

Design details

Randomization method
Computer-based stratified randomization implemented through the online survey platform. For the information treatment experiment, participants are assigned to one of three video conditions (Treatment 1, Treatment 2, or Control) using a stratified randomization algorithm that ensures balance across key demographic characteristics. The randomization procedure maintains approximately equal distribution of age groups (19-24, 25-29, 30-34, 35-39), gender, and regional residence (metropolitan versus non-metropolitan) across all three experimental conditions. Within each stratum, participants are randomly assigned to treatment ensuring both randomization and demographic balance across treatment groups.
Randomization unit
Individual participant level with stratification. Each survey respondent is independently randomized to treatment conditions within their demographic stratum defined by age group, gender, and regional residence.
Clusters
N/A. individual level randomization
Observations
3,200 individual participants aged 19-39 years
Arms
[Information Treatment Experiment] Approximately 1,067 participants in Treatment 1 (factual housing cost video), 1,067 participants in Treatment 2 (AI dialogue video), and 1,066 participants in Control (tourism placebo video), with balanced representation of age groups, gender, and regional residence across all three arms. All 3,200 participants complete the DCE component. Regional oversample: The sample includes an oversample of 1,000 participants from Daegu and North Gyeongsang Province (approximately 33% of total sample) to ensure robust analysis for this region given Pohang's strategic importance for regional development policy.
Primary outcomes
[Discrete Choice Experiment Outcomes] We collect binary choice indicators from eight paired comparisons in the general module and eight paired comparisons in the Pohang-specific module, recording which workplace option (1 or 2) participants select in each scenario. From these choices, we estimate preference parameters using conditional logit or mixed logit models to determine coefficients for each job and location attribute, calculate willingness-to-pay estimates as ratios of attribute coefficients to the income coefficient, and examine heterogeneous preferences across demographic subgroups including age, current residence, and regional experience. [Information Treatment Experiment Outcomes] We measure participants' intention to work and reside in non-metropolitan regions through Question P1, using a 5-point Likert scale ranging from no intention at all to strong intention. This outcome is measured immediately following exposure to the treatment videos.
Secondary outcomes
[Sector-Specific Regional Work Intentions] We measure willingness to work in regional public institutions through Question P2 and regional large corporations through Question P3, using the same 5-point scale as the primary outcome. [Information Treatment Effects on Housing Cost Perceptions] We assess changes in perceived housing purchase burden indices for five regional cities (Gyeonggi, Gwangju, Daejeon, Daegu, and Ulsan), calculated as the difference between post-treatment responses in Question P5 and baseline responses in Question A4. [Information Sharing Intentions] We examine participants' likelihood of sharing video information with others through Question P4, which uses a 5-point scale to measure the perceived credibility and usefulness of the treatment information. [Subgroup-Specific Treatment Effects on Regional Work Intentions] We measure the same regional work intention outcome (Question P1) separately for demographic subgroups including metropolitan versus regional residents, those with and without 6+ months of prior regional residence experience, four age cohorts (19-24, 25-29, 30-34, 35-39), and employed versus non-employed participants.

Cited by

No citation data available for this registration yet.

Keywords