An exploratory replication and falsification study of migrant life-satisfaction convergence.
Publication type: Research Report
Version: 1.0
Date: 30 September 2026
Institution: The Null Institute
Author: Ilya Nikitin, Ph.D., Founder and Director, The Null Institute
Status: Exploratory observational reanalysis; not peer reviewed
Abstract
Background. Previous migration research has shown that migrants’ life evaluations tend to move toward those of destination populations. The existence of this destination-convergence pattern is therefore not novel. A narrower unresolved question is how much of the observed convergence reflects changes in life evaluation itself and how much may arise because migrants adopt the destination population’s way of using numerical rating scales.
Objective. To replicate destination convergence in European Social Survey data and then attempt to falsify it using response-style measures derived from other survey items.
Methods. We analysed pooled European Social Survey rounds 1–11, matching migrants with native-born respondents in their countries of origin and destination within the same survey rounds. A stricter analysis retained adults who migrated after 1991. For a core adult post-1991 specification, destination transfer in life satisfaction was estimated from 20 origin-destination routes involving 2,361 migrants. We separately quantified response behaviour using 19 other 0–10 ESS items, excluding life satisfaction and happiness, and tested midpoint and endpoint heaping as well as within-person transformations of the main wellbeing measures.
Results. In the core adult post-1991 analysis, the life-satisfaction transfer slope was 0.662 with a route-bootstrap 95% interval of 0.564–0.844. ESS global happiness produced a similar slope of 0.661 (95% interval 0.508–0.996). Migrants moving toward lower-life-satisfaction destinations also tended to shift downward relative to origin natives, weakening a simple “positive people migrate” explanation. Convergence did not increase monotonically with years since migration.
Response style itself also converged toward the destination. In preserved summary models, midpoint-related response behaviour showed moderate destination alignment, while endpoint-related behaviour showed stronger alignment. However, life-satisfaction destination convergence remained after alternative within-person centring, standardisation and rank-based transformations. Response-style adaptation therefore exists, but in these data it does not eliminate the destination signal in life satisfaction.
Conclusions. The study does not identify a causal effect of migration. Selection into migration, destination choice, changing reference groups and selective return remain unresolved. The narrower finding is that migration changes both reported life evaluation and aspects of the numerical response process used to express it, while measured response-style adaptation is insufficient to explain the full observed convergence.
1. Research question
Country happiness rankings are descriptive. They tell us how residents of different countries answer questions about their lives. A potential migrant asks a more difficult question: if I move from country A to country B, will my evaluation of life move toward the level reported in country B?
That question is not answered by a national ranking alone. Cross-country differences may reflect institutions, safety, employment, income, health and social relations, but also selection, culture, language, reference groups or response style.
Migration provides a useful stress test because a person originates in one environment and is later observed in another. Previous work already shows substantial destination convergence, including analyses in the World Happiness Report 2018 and the study by Helliwell, Shiplett and Bonikowska on migrants to Canada and the United Kingdom. Our contribution is therefore not to claim discovery of destination convergence. We ask whether the pattern survives a direct measurement challenge:
Could migrants appear destination-like mainly because they learn to use rating scales the way destination natives do?
2. Data and design
European Social Survey
The primary data source was the European Social Survey, rounds 1–11. The pooled source file contained approximately 541,000 respondents, around 50,000 of whom were born outside the country where they were interviewed.
For each migration route we sought three groups observed within matched ESS rounds:
- native-born respondents living in the origin country;
- respondents born in that origin country but living in the destination;
- native-born respondents living in the destination.
Matching within ESS round reduces a major source of bias: comparing migrants interviewed in one period with origin or destination populations measured in a different historical period.
Transfer coefficient
For descriptive interpretation we define:
T = (M − O) / (D − O)
where O is the origin-native mean, M the migrant mean, and D the destination-native mean.
T = 0 means the migrant mean equals the origin-native mean. T = 1 means it equals the destination-native mean. Values outside 0–1 are possible. This is not a causal fraction and must not be read as “the percentage of happiness caused by the country”. The ratio is also unstable when the native origin-destination gap is small, so aggregate slope estimates are preferred to naïve averaging of route ratios.
Strict migration specification
A broad foreign-born definition can mix very different historical processes. One example emerged in Russia → Estonia comparisons, where many Russian-born respondents had moved before 1991, when the relevant international border did not exist in its present form.
We therefore used a stricter specification for key analyses:
- migration after 1991;
- age at migration at least 18 years.
3. Replication of destination convergence
In the adult post-1991 mechanism dataset, life satisfaction showed a destination transfer slope of 0.662 across 20 migration routes involving 2,361 migrants. The route-bootstrap 95% interval was 0.564–0.844.
ESS global happiness showed a very similar transfer slope of 0.661 across 20 routes and 2,368 migrants, with a wider 95% interval of 0.508–0.996.
| Outcome | Routes | Migrants | Transfer slope | 95% bootstrap interval |
|---|---|---|---|---|
| Life satisfaction | 20 | 2,361 | 0.662 | 0.564–0.844 |
| ESS global happiness | 20 | 2,368 | 0.661 | 0.508–0.996 |
These results are consistent with earlier international migration research rather than replacing it. Their value here is that the same ESS construction can be subjected to additional falsification tests.
4. Reverse-direction test
A simple positive-selection story predicts that migrants may appear happier than origin residents because unusually optimistic, healthy or successful people are more likely to migrate.
One falsification is to examine moves toward destinations whose native residents report lower life satisfaction than residents of the origin country.
For life satisfaction, the estimated transfer slope was approximately 0.82 when the destination-native mean was higher and 0.53 when it was lower. For ESS global happiness, the corresponding slopes were approximately 0.88 and 0.29.
The downward-moving samples were much smaller and the estimates less precise. Nevertheless, the direction is informative: migrants did not show a universal upward shift. When the destination scored lower, migrant means tended to move downward as well.
This does not remove sophisticated selection mechanisms. Different people choose different destinations. It does weaken the simplest explanation in which migrants look destination-like merely because migrants are unusually positive people.
5. Years since migration
If destination convergence were a slow monotonic assimilation process, one might expect the transfer coefficient to increase steadily with time spent in the destination.
That pattern was not observed.
| Years since migration | Life-satisfaction transfer slope |
|---|---|
| 0–5 | 0.925 |
| 6–10 | 0.842 |
| 11–20 | 0.724 |
| 21+ | 0.874 |
Global happiness showed the same broad absence of a monotonic rise.
This does not prove rapid causal adaptation. ESS is mainly repeated cross-sectional data, not a panel following the same migrants for decades. Different tenure groups can differ in migration cohort, admission regime, age structure and selective return. The narrower point is that the observed destination alignment is already substantial among relatively recent migrants and does not simply grow smoothly with duration of residence.
6. What else moves toward the destination?
Using the same adult post-1991 framework, several additional measures also moved toward destination-native levels.
| Measure | Transfer slope | 95% bootstrap interval |
|---|---|---|
| Institution/economy evaluations | 1.20 | 1.09–1.37 |
| Feeling safe after dark | 1.17 | 0.92–1.40 |
| Self-rated health | 0.94 | 0.69–1.29 |
| Social trust | 0.78 | 0.60–0.91 |
| Life satisfaction | 0.66 | 0.56–0.84 |
| Income comfort | 0.52 | 0.25–0.84 |
| Frequency of social meetings | 0.43 | 0.22–0.70 |
These patterns are descriptive. In particular, evaluations of institutions, health and income comfort are themselves subjective responses. They should not be treated as identified causal mediators.
7. The response-style falsification
This was the most important test.
International surveys are vulnerable to response-style differences. Some populations use scale endpoints readily; others avoid them. Some prefer the midpoint. If migrants adopt the numerical conventions of the destination, life satisfaction could appear to converge even if lived experience changed less.
We therefore constructed response-style measures from 19 other ESS 0–10 items, excluding life satisfaction and happiness.
Response style itself moves
The falsification did not produce the convenient answer that response behaviour is fixed.
Migrants did become more destination-like in how they used rating scales. In preserved summary models:
- midpoint-related heaping showed moderate destination alignment, approximately 0.48–0.52 depending on specification;
- endpoint-related heaping showed stronger destination alignment, around 0.70 in the intercept model;
- broader endpoint/focal-response measures also showed substantial convergence.
Thus response style is not merely a stable imprint of the origin culture. It can itself adapt after migration.
But life-satisfaction convergence remains
The core question is whether removing general scale-use tendencies causes the life-satisfaction result to disappear.
It did not.
In the preserved response-style analysis, raw life-satisfaction destination alignment was approximately 0.73. Alternative within-person transformations were then applied using the respondent’s answers to the other 0–10 items.
A within-person standardised life-satisfaction specification retained an intercept-model destination slope close to 0.73. Rank-based normalisation weakened the signal but left it destination-oriented at approximately 0.67 in the corresponding intercept model.
These transformations are not measurements of “true happiness after correcting error”. The auxiliary questions contain real substantive information, so removing common response patterns can remove genuine adaptation together with stylistic scale use.
The defensible conclusion is narrower:
Response-style assimilation exists, but the response-style changes measured here are not sufficient to account for the full destination alignment observed in life satisfaction.
8. External replication outside the ESS setting
We also reanalysed the regional migration-flow table published with the World Happiness Report 2018.
Across all 20 published flows, the slope linking migrant life-evaluation gain to the native destination-origin gap was approximately 0.60 unweighted and 0.70 weighted by migrant sample size.
Restricting to inter-regional moves gave approximately 0.55 unweighted and 0.61 weighted. The native-gap to migrant-gain correlation among those inter-regional flows was approximately 0.83.
Examples include:
| Migration flow | Native gap | Migrant gain | Gain/gap ratio |
|---|---|---|---|
| Sub-Saharan Africa → Western Europe | +2.29 | +1.44 | 0.63 |
| MENA → Western Europe | +1.67 | +0.90 | 0.54 |
| Central/Eastern Europe → Western Europe | +1.16 | +0.78 | 0.67 |
| CIS → Western Europe | +1.28 | +0.59 | 0.46 |
| South Asia → Southeast Asia | +1.71 | +0.80 | 0.47 |
The numerical similarity should not be overinterpreted. WHR and ESS use different sampling frames and geographic constructions. What matters is that substantial destination alignment is not confined to one European dataset.
9. What stronger designs say
Observational migration data cannot identify the counterfactual outcome for each migrant. Stronger designs therefore matter.
Helliwell, Shiplett and Bonikowska reported destination convergence among immigrants to Canada and the United Kingdom, including distributional evidence and comparisons of similar origin groups living in different destinations.
A longitudinal study of Ingrian Finnish migrants from Russia to Finland measured the same individuals before and after migration and found increasing life satisfaction, while self-esteem followed a different trajectory.
The Tonga → New Zealand migration lottery provides an important warning against compressing wellbeing into one outcome. Migration improved several objective and mental-health measures while one direct happiness measure declined.
The appropriate conclusion is not “migration to a happier country makes people happier”. Different dimensions of wellbeing can move in different directions.
10. Interpretation
Life satisfaction and Cantril-type life evaluation are often criticised because they are sensitive to external circumstances, social comparison and ideas about what a good life should look like.
For a philosophical measure of context-free inner happiness, that sensitivity may indeed be a limitation.
For migration, however, sensitivity to environment is potentially useful. A prospective migrant is specifically interested in the part of wellbeing that can change when the surrounding conditions change.
We therefore use the phrase environment-sensitive signal rather than claiming to have identified a latent quantity such as “transferable quality of life”. National life evaluation appears to contain information about group-level differences that migrants partly acquire after relocation.
This is not a statement that a country determines a fixed percentage of anyone’s happiness.
11. Limitations
- Selection into migration. Migrants are not randomly drawn from origin populations.
- Destination selection. People choose destinations, and different destinations attract different migrants.
- Selective return. Migrants whose experience goes badly may leave and disappear from later destination samples. ESS cannot resolve this.
- Cross-sectional tenure groups. Years-since-migration comparisons do not follow the same individuals through time.
- Reference-group adaptation. A genuine change in answers may reflect a changed comparison standard as well as changed circumstances.
- Response-style measurement is imperfect. The auxiliary scales have substantive content and therefore cannot provide a pure instrument for numerical style.
- Dependency structure. Migration routes share origins and destinations, and repeated ESS rounds reuse country environments. The present route-bootstrap does not fully model all multiway dependence.
- Exploratory analysis. The research was not preregistered. Multiple hypotheses and falsification tests were developed sequentially as the investigation progressed.
- Group means are not individual predictions. A route-level transfer coefficient cannot predict what will happen to a particular migrant.
12. Reproducibility status
The principal findings reported here are backed by preserved derived artifacts:
- bilateral ESS origin–migrant–destination results;
- adult post-1991 migration results;
- tenure summaries;
- reverse-direction tests;
- migration-mechanism transfer summaries;
- migration-mechanism regressions;
- WHR 2018 regional-flow reanalysis;
- item-level response-style/heaping data;
- response-style falsification summaries;
- within-person response-style invariant summaries.
An additional exploratory analysis using 101 route-by-round observations across 34 migration routes was recorded in the research log, but its standalone final derived artifact was not retained. That analysis is therefore intentionally excluded from the main results of this report.
Before journal submission, the highest-value technical extensions would be:
- leave-one-origin-out and leave-one-destination-out sensitivity analyses;
- a multiway or hierarchical uncertainty model accounting for shared origins, shared destinations and ESS rounds;
- an independently rerun executable pipeline from ESS source data to every table;
- a confirmatory replication on a dataset not used during hypothesis development.
13. Conclusion
Destination convergence in migrant life satisfaction is not a new phenomenon. This study reproduces it in ESS data and then subjects it to a measurement-based falsification test.
The falsification partly succeeds: migrants do adopt aspects of the destination population’s numerical response style. Midpoint and endpoint use are not fixed cultural fingerprints.
But the stronger claim fails. Adjusting life satisfaction relative to broader individual scale use does not make the destination signal disappear.
The resulting picture is layered:
- environment changes;
- social and institutional perceptions change;
- reference groups may change;
- numerical response habits change;
- and broad evaluation of life changes with them.
Migration changes both reported life evaluation and the response process used to express it. In these data, response-style adaptation is measurable but does not account for the full observed destination convergence.
The main unresolved question is causal decomposition: how much of the remaining convergence reflects changed material conditions, institutions, social environment, reference standards, selective migration and selective return. A strong next design would follow the same people before migration, repeatedly after migration and after any return migration, while measuring life evaluation, experienced affect and response style separately.
References
- European Social Survey. ESS data portal, rounds 1–11.
- Helliwell JF, Huang H, Wang S, Norton M. World Happiness Report 2018, Chapter 2–3 and migration appendix materials.
- Helliwell JF, Shiplett H, Bonikowska A. Migration as a test of the happiness set-point hypothesis: Evidence from immigration to Canada and the United Kingdom. Canadian Journal of Economics. 2020.
- Lönnqvist J-E et al. The mixed blessings of migration: Life satisfaction and self-esteem over the course of migration. European Journal of Social Psychology. 2015.
- Stillman S, Gibson J, McKenzie D, Rohorua H. Miserable Migrants? Natural Experiment Evidence on International Migration and Objective and Subjective Well-Being. World Development. 2015;65:79–93.
- Blanchflower DG, Bryson A. Wellbeing Rankings. Social Indicators Research. 2024.
Research provenance
The analysis was conducted as part of The Null Institute’s independent research program. AI tools assisted source recovery, data handling, falsification testing, checking and drafting under human direction. Responsibility for the published interpretation remains with the named author.