Exit and Voice under Different Institutional Frames: Evidence from Job-Search and Collective Bargaining Survey Experiments

This study examines how institutional framing affects the evaluation of employment conditions and behavioural responses in the labour market. Participants are randomly assigned to one of two institutional frames that describe identical employment conditions in different ways. In the job-search framing, employment conditions are presented as individual job offers that workers encounter when searching for a new job. In the collective bargaining framing, the same employment conditions are presented as outcomes of negotiated tariff agreements between employers and unions.

Across both frames, employment packages vary along four core attributes: monthly gross income, weekly working hours, vacation days per year, and employment guarantees against dismissal. Attribute levels are expressed relative to respondents’ current employment conditions in order to make the scenarios realistic and comparable to common labour market changes.

Participants complete several survey-based experimental tasks. In a discrete choice experiment, respondents repeatedly choose between hypothetical employment packages that differ in their attribute levels. In additional factorial survey experiments, respondents evaluate hypothetical scenarios describing employment conditions and report behavioural intentions such as willingness to participate in a strike or likelihood of accepting an alternative job offer.

The experimental variation of employment attributes allows the study to estimate how workers value different job characteristics and how these valuations translate into behavioural responses such as job exit or participation in collective action.

The study investigates how institutional framing influences the evaluation of employment conditions and how these evaluations translate into behavioural responses in the labour market. The intervention consists of experimentally varying the institutional context in which employment conditions are presented, while holding the underlying economic attributes constant.

Participants are randomly assigned to one of two institutional frames. In the job-search framing, employment conditions are presented as outcomes of individual job search, where workers compare alternative job offers. In the collective bargaining framing, the same employment conditions are described as negotiated outcomes of collective bargaining agreements between employers and unions. This design allows the study to isolate how institutional context affects the valuation of job attributes and behavioural responses.

Across both treatments, employment packages vary along four attributes: monthly gross income, weekly working hours, vacation days per year, and employment guarantees against dismissal. Attribute levels are expressed relative to respondents’ current employment situation to ensure realistic and meaningful comparisons. These attributes are varied experimentally across tasks.

The empirical design consists of three complementary survey-based experiments.

Experiment 1 uses a discrete choice experiment (DCE) to study how institutional framing affects wage–amenity trade-offs. Respondents repeatedly choose between hypothetical employment packages that vary along the four employment attributes. The framing treatment allows comparison of attribute valuations when employment conditions are interpreted as job offers versus negotiated tariff outcomes.

Experiment 2 employs a factorial survey experiment in the collective bargaining treatment to analyse behavioural responses within a tariff context. Respondents evaluate hypothetical tariff agreements that differ in the four employment attributes and report their willingness to support collective action, such as participating in a strike.

Experiment 3 examines the relationship between exit and voice responses within individuals. Respondents evaluate hypothetical job opportunities and report their likelihood of accepting a job offer (exit) as well as their willingness to participate in collective action (voice). By presenting identical economic conditions across behavioural contexts, the experiment allows assessment of whether exit and voice function as substitute or complementary behavioural strategies.

The experimental design therefore links preference-based valuations of employment conditions with behavioural responses across different institutional contexts. By combining discrete choice and factorial survey methods, the study provides evidence on how workers translate changes in employment conditions into labour market behaviour.

2026-03-25

2026-04-15

The primary outcomes correspond to respondents’ choices and behavioural intentions across the three experimental modules.

In Experiment 1 (Discrete Choice Experiment), the primary outcome is the choice between alternative employment packages that differ in income, working hours, vacation days, and employment guarantees. These employment packages are presented either under a job-search framing, where they represent individual job offers, or under a collective bargaining framing, where they represent negotiated tariff outcomes. The outcome variable records which employment package respondents choose in each decision task.

In Experiment 2 (Tariff Factorial Survey Experiment), the primary outcome is strike willingness. Respondents evaluate hypothetical tariff agreements that differ in employment conditions and report their willingness to support collective action such as participating in a strike.

In Experiment 3 (Exit–Voice Factorial Survey Experiments), the primary outcomes are exit and voice behavioural intentions. Exit is measured as the stated likelihood of accepting an alternative job offer under different employment conditions. Voice is measured as the stated willingness to participate in collective action in response to changes in employment conditions.

Together, these outcomes capture how workers evaluate employment conditions and how these evaluations translate into behavioural responses across different institutional contexts.

The primary outcome variables are measured using respondents’ choices and behavioural ratings in the experimental tasks.

In Experiment 1 (Discrete Choice Experiment), the outcome variable records which employment package respondents select in each choice task. Each task presents two hypothetical employment packages that differ in income, working hours, vacation days, and employment guarantees. The dependent variable is coded as a binary indicator identifying the selected alternative. The experimental design allows estimation of how variation in these attributes affects choice probabilities under different institutional frames.

In Experiment 2 (Tariff Factorial Survey Experiment), the primary outcome is strike willingness. Respondents evaluate hypothetical tariff agreements that differ in employment conditions and report their willingness to support collective action. Strike willingness is measured using a rating scale capturing respondents’ support for participating in industrial action under the described conditions.

In Experiment 3 (Exit–Voice Factorial Survey Experiments), two behavioural outcomes are measured. Exit corresponds to the stated likelihood of accepting an alternative job offer under the described employment conditions. Voice corresponds to the stated willingness to participate in collective action, such as a strike, in response to changes in employment conditions. Both outcomes are measured using rating scales that capture respondents’ behavioural intentions under each vignette.

Across the factorial survey experiments, respondents evaluate multiple scenarios, allowing behavioural responses to be analysed as a function of experimentally varied employment attributes.

The study also collects several secondary outcomes that capture respondents’ perceptions of the experimental tasks and their interpretation of the decision context.

These secondary outcomes include:

Perceived realism of the experimental tasks, measuring whether respondents considered the presented scenarios realistic and comparable to real-life employment decisions.

Collective orientation in decision-making, capturing whether respondents interpreted the decision context primarily as an individual employment decision or as part of a collective labour relations context.

These outcomes are measured using post-experimental survey items administered after the experimental tasks.

The secondary outcomes are constructed from post-experimental survey items that measure respondents’ perceptions of the experimental tasks and their interpretation of the decision context.

The Perceived Realism Index captures whether respondents considered the experimental scenarios realistic and behaviourally meaningful. It is constructed as the standardized mean of three Likert-scale items asking respondents whether (i) the scenarios appeared realistic, (ii) the described employment conditions seemed plausible, and (iii) they approached the decisions in a way similar to real-life employment decisions. Higher values indicate greater perceived realism.

The Collective Orientation Index measures whether respondents interpreted the decision context as individual or collective in nature. The index is constructed as the standardized mean of three items that capture whether respondents considered collective outcomes, collective bargaining logic, or individual job decisions when evaluating the scenarios. One item is reverse-coded so that higher values consistently indicate a stronger collective interpretation of the decision context.

Both indices are standardized to facilitate interpretation and comparison across respondents.

The study consists of three survey-based experiments designed to examine how workers evaluate employment conditions and how these evaluations translate into behavioural responses in the labour market.

Participants are randomly assigned to one of two institutional frames. In the job-search framing, employment conditions are presented as individual job offers encountered during job search. In the collective bargaining framing, identical employment conditions are described as negotiated outcomes of collective bargaining agreements between employers and unions. The framing treatment allows the study to examine whether institutional context affects how workers evaluate employment conditions.

Across treatments, employment packages vary along four attributes: monthly gross income, weekly working hours, vacation days per year, and employment guarantees against dismissal. Attribute levels are expressed relative to respondents’ current employment situation. These attributes are experimentally varied across tasks.

The experimental design consists of three modules. Experiment 1 is a discrete choice experiment in which respondents repeatedly choose between alternative employment packages. Experiment 2 is a factorial survey experiment in the collective bargaining framing that measures respondents’ willingness to support collective action, such as participating in a strike, under different tariff agreements. Experiment 3 consists of factorial survey experiments in the job-search framing that measure respondents’ behavioural intentions along the exit and voice margins, including the likelihood of accepting alternative job offers and the willingness to participate in collective action.

Because employment attributes are experimentally varied across decision tasks, the design allows causal estimation of how economic conditions and institutional framing affect both preference formation and behavioural responses.

Overview of experimental design

The study combines discrete choice and factorial survey methods to examine how workers evaluate employment conditions and how these evaluations translate into behavioural responses in the labour market. The experimental design links preference elicitation with behavioural intentions across different institutional contexts.

Institutional framing treatment

Participants are randomly assigned at the individual level to one of two institutional frames. In the job-search framing, employment conditions are presented as individual job offers that workers encounter when searching for a new job. In the collective bargaining framing, the same employment conditions are described as negotiated outcomes of collective bargaining agreements between employers and unions. This design allows the study to isolate the causal effect of institutional framing while holding the underlying employment attributes constant.

Across both frames, employment packages vary along four experimentally manipulated attributes: monthly gross income, weekly working hours, vacation days per year, and employment guarantees against dismissal. Attribute levels are expressed relative to respondents’ current employment conditions to ensure that the scenarios resemble realistic changes in employment conditions observed in job mobility and collective bargaining agreements.

Experimental modules

The empirical design consists of three complementary experimental modules.

Experiment 1: Discrete Choice Experiment

Respondents participate in a discrete choice experiment in which they repeatedly choose between two hypothetical employment packages. Each employment package is defined by the four employment attributes described above. The framing treatment determines whether these employment packages are interpreted as individual job offers (job-search framing) or as negotiated tariff outcomes (collective bargaining framing). The experiment allows the estimation of how institutional framing affects the valuation of job attributes and wage–amenity trade-offs.

Experiment 2: Tariff Factorial Survey Experiment

In the collective bargaining treatment, respondents evaluate hypothetical tariff agreements that vary along the same employment attributes. For each scenario, respondents report their willingness to support collective action such as participating in a strike. This experiment analyses how employment conditions affect strike willingness within a collective bargaining context.

Experiment 3: Exit–Voice Factorial Survey Experiments

In the job-search treatment, respondents evaluate hypothetical job opportunities and report their behavioural intentions along two dimensions. First, respondents indicate the likelihood that they would accept the job offer (exit). Second, respondents indicate their willingness to participate in collective action such as a strike (voice). Presenting identical employment conditions across these decision contexts allows the analysis of whether exit and voice function as substitute or complementary behavioural strategies.

Across the factorial survey experiments, respondents evaluate multiple scenarios, allowing behavioural responses to be analysed as a function of experimentally varied employment attributes. Because employment conditions are randomly varied across experimental tasks, the design allows causal identification of how economic incentives and institutional framing affect both preferences and behavioural intentions.

Pre-specified Hypotheses

The study tests several pre-specified hypotheses regarding the valuation of employment conditions and behavioural responses to changes in working conditions.

H1 (Preference for improved employment conditions). Improvements in employment attributes — including higher income, fewer working hours, more vacation days, and stronger employment protection — increase the probability that an employment package is chosen in the discrete choice experiment.

H2 (Loss aversion around the employment status quo). Negative deviations from current employment conditions in non-wage attributes — such as increases in working hours or reductions in vacation days — exert stronger effects on utility than comparable improvements.

H3 (Institutional framing effects). The valuation of employment attributes differs between the job-search framing and the collective bargaining framing.

H4 (Perceived decision context). Institutional framing affects respondents’ perceptions of the decision context, including perceived realism of the task, behavioural consequentiality, and collective decision orientation.

H5 (Strike willingness and employment conditions). In the tariff factorial survey experiment, improvements in employment conditions increase respondents’ willingness to support collective action such as strikes.

H6 (Exit–voice differentiation). The effects of employment attributes differ across exit and voice responses when individuals evaluate identical economic conditions.

H7 (Substitution between behavioural strategies). Individuals with a higher baseline propensity for exit exhibit a lower baseline propensity for voice.

Hypotheses and empirical tests

The hypotheses will be evaluated using the experimental variation in employment attributes and institutional framing across the three experimental modules.

H1 (Preference for improved employment conditions). Tested in the discrete choice experiment by estimating the effects of income increases, reductions in working hours, increases in vacation days, and longer employment guarantees on the probability that an employment package is chosen.

H2 (Loss aversion around the employment status quo). Tested by comparing the magnitude of coefficients associated with negative and positive deviations from current employment conditions in non-wage attributes.

H3 (Institutional framing effects). Tested using interaction terms between employment attributes and the institutional framing treatment in the discrete choice experiment.

H4 (Perceived decision context). Tested by estimating treatment-effect regressions in which indices of perceived realism, behavioural consequentiality, and collective decision orientation are regressed on the institutional framing treatment indicator. As a behavioural complement, we test for treatment-specific scale differences in a mixed logit model.

H5 (Strike willingness). Tested in the tariff factorial survey experiment by estimating how experimentally varied employment conditions affect reported strike willingness.

H6 (Exit–voice differentiation). Tested by comparing attribute effects across exit and voice responses within individuals using interaction terms between behavioural context (exit vs. voice) and employment attributes.

H7 (Substitution between behavioural strategies). Tested by examining the relationship between baseline exit and voice propensities across respondents.

Pre-specified exploratory analyses

In addition to the confirmatory hypotheses, the study includes several pre-specified exploratory analyses intended to examine relative effect magnitudes and potential heterogeneity underlying the main results.

First, the analysis will explore the relative importance of different employment attributes across experimental modules, including comparisons between income and non-wage attributes in the discrete choice experiment and the salience of tariff attributes in shaping strike willingness.

Second, the analysis will examine whether the valuation of employment attributes and behavioural responses differ across worker characteristics such as income level, sector of employment (public versus private), union membership, and occupational group.

Third, exploratory analyses will provide descriptive evidence on variation in exit and voice propensities across individuals and behavioural contexts.

These analyses are considered exploratory and will be interpreted as suggestive evidence rather than confirmatory tests.

Linking preference valuations and behavioural responses

The experimental design intentionally links preference measurement and behavioural responses across experimental modules. The discrete choice experiment measures how respondents value employment attributes in terms of wage–amenity trade-offs. The factorial survey experiments then examine how these employment conditions translate into behavioural intentions such as strike participation and job mobility.

Using identical employment attributes across experimental modules allows comparison between preference-based valuations and behavioural responses under different institutional contexts. This design makes it possible to examine whether employment attributes that are highly valued in the choice experiment also induce stronger exit or voice responses in the factorial survey experiments.

Identification strategy

Causal identification in the study relies on two sources of experimental variation. First, respondents are randomly assigned to one of two institutional framing conditions, allowing estimation of how the institutional context in which employment conditions are presented affects the valuation of job attributes and behavioural responses.

Second, employment attributes—including income, working hours, vacation days, and employment guarantees—are experimentally varied across decision tasks in both the discrete choice and factorial survey experiments. This variation allows estimation of how changes in employment conditions affect respondents’ choices and behavioural intentions.

Because both the institutional framing and the employment attributes are randomly assigned, estimated effects can be interpreted as causal effects of institutional context and employment conditions on respondents’ preferences and behavioural intentions under standard assumptions of experimental survey designs.

The pre-analysis plan accompanying this registration specifies the econometric models and statistical tests used to evaluate the hypotheses.

Randomization is implemented by the survey software using computer-generated random assignment. Respondents are randomly assigned to either the job-search framing or the collective bargaining framing with equal probability. The assignment occurs automatically when respondents enter the experimental module of the survey.

Randomization occurs at the individual respondent level. Each respondent is assigned to one of the two institutional framing conditions.

No

Approximately 1,800 individual respondents.

Approximately 1,800 respondents completing the experimental survey.

Approximately 900 respondents in the job-search framing and 900 respondents in the collective bargaining framing.

A formal ex-ante power calculation is not conducted because the experimental design combines multiple survey experiments with repeated observations per respondent, which substantially increases statistical power relative to simple between-subject designs. The target sample size of approximately 1,800 respondents (around 900 per treatment condition) is chosen to ensure sufficient statistical power to detect moderate treatment effects across the experimental modules.