“Migration” is often used as if it names one measurable thing. It does not. It can mean moving across a border, having foreign citizenship, being born abroad, having one or two parents born abroad, seeking asylum, arriving for work or study, moving within the EU, joining family, or being read by others as not belonging.
That category problem is not a technical footnote. It shapes nearly every claim made in public debate. The question is not only whether a statement is true. It is: true for which group, in which data source, over which period, compared with whom, and with what uncertainty?
Myth: “The data clearly proves whether migration is good or bad.”
Reality: data can answer narrower questions. It cannot deliver a single moral verdict.
Research can examine employment, education, housing, health, public finances, language access, discrimination, political participation, and social trust. Some findings are robust, some are mixed, and some depend strongly on policy design. The answer also changes depending on whether the focus is the person who moved, the receiving society, the country of origin, employers, public budgets, schools, or local communities.
The supported claim is that migration has multiple effects, not one effect. Labour-market outcomes differ by recognition of qualifications, residence status, language opportunity, discrimination, childcare access, local job demand, and time since arrival. Education outcomes differ by school system, segregation, family resources, language support, and whether children migrated themselves or were born in Germany.
The misleading claim is any sentence that treats “migration” as one lever producing one result. A serious evidence discussion needs the mechanism: what pathway is being claimed, and what evidence connects migration to that pathway?
Myth: “If two official sources disagree, one must be lying.”
Reality: sources often count different things for different purposes.
Migration statistics vary because administrative systems and surveys define populations differently. A residents’ register may count registered addresses. Asylum statistics may count applications, decisions, or protection statuses. Labour-market statistics may count employment status, citizenship, or country of birth depending on the table. The Mikrozensus can describe households and migration background using survey categories. SOEP panel data can follow people over time and connect migration-related variables to income, work, family, and attitudes.
These are not interchangeable. A person may be a German citizen and still have a migration background. Someone may be foreign-born but long settled. Someone may be newly arrived but not represented clearly in a survey sample. People without stable registration can be missing from many systems.
Source differences become dangerous when a number is pulled out of its original purpose. A service statistic about people who came through one advice centre cannot describe all migrants in Germany. A survey of adults cannot automatically describe children. A table on nationality cannot answer a question about racialization. A count of asylum applications cannot describe all migration.
Good interpretation begins with the boring question: who is included in the denominator?
Myth: “Migration background tells us what we need to know.”
Reality: migration background is useful for some questions and blunt for others.
In Germany, “Migrationshintergrund” has been used to describe whether a person or their parents migrated under particular definitions. It can reveal inequalities that would be hidden if only current citizenship were measured. For example, it can help researchers see whether disadvantages persist among people born in Germany.
But the category groups together people with very different histories, languages, legal statuses, class positions, racialized experiences, religions, and lengths of residence. It may include a recently arrived refugee, a German-born university student with one foreign-born parent, an EU worker, and a naturalized pensioner. Their social positions are not the same.
The supported claim is that broad categories can show broad inequalities. The contested part is interpretation. If an outcome differs by migration background, the category itself does not explain why. The difference may involve discrimination, parental education, wealth, school tracking, residence status, language support, neighbourhood segregation, labour-market access, or measurement bias.
The claim we cannot make is that migration background is a cause by itself. It is a flag for further analysis, not a diagnosis.
Myth: “Personal stories are enough to prove the general trend.”
Reality: stories can identify mechanisms and harms, but they cannot estimate scale by themselves.
A testimony about a qualification not being recognized, a school assuming lower ability, or a landlord reacting differently to a surname can be important evidence of how systems are experienced. Interviews and ethnographic research can show what a survey question would miss: fear, adaptation, informal rules, humiliation, or the sequence of small barriers that make a formal right hard to use.
But one story cannot tell us how common the pattern is. For that, researchers need surveys, administrative data, audit studies, or other systematic designs. The reverse is also true: a dataset can show a gap but fail to explain how it feels or how people navigate it.
The responsible position is not stories versus statistics. It is matching evidence to claim. Use stories to understand mechanisms, meanings, and consequences. Use systematic data to estimate distribution, change over time, and differences between groups. Use both to avoid a public debate that is either abstract or anecdotal.
Community Voices Media is the stronger lane for longform testimony and media craft. The Institute’s lane is to explain what kind of claim each form of evidence can carry.
Myth: “If we cannot measure something perfectly, we should not talk about it.”
Reality: imperfect evidence can still be informative if its limits are named.
Some migration-related questions are difficult to measure because categories are sensitive, trust is low, or administrative systems were not built for research. Racial discrimination is a clear example. Germany has limited official data on racialization compared with some other contexts. Nationality or migration background may capture part of the picture, but they do not measure how a person is perceived in a housing viewing, school meeting, or police encounter.
Undocumented or precariously housed people may be missing from official data. People with limited German or low trust in institutions may be underrepresented in surveys. Small subgroups may be statistically invisible because sample sizes are too low. Data protection, including GDPR principles, rightly limits careless collection of sensitive information.
The supported claim is that gaps exist and should be named. The misleading claim is that absence from a dataset means absence of harm. Evidence gaps should make us more careful, not silent.
Myth: “A single trend explains integration.”
Reality: integration is a bundle of outcomes, and trends can move in different directions.
“Integration” can refer to language acquisition, employment, education, health access, civic participation, social networks, legal security, income, housing, or belonging. Some indicators may improve while others stagnate. A person may speak fluent German and still face discrimination. A family may be economically stable but excluded from local decision-making. A child may do well in school while parents remain isolated from institutions.
Panel data, which follows the same people over time, is especially useful here. It can show trajectories: how outcomes change with time since arrival, education, family changes, policy shifts, and labour-market entry. Cross-sectional data, which captures a population at one point in time, is useful too, but it can confuse cohort differences with individual change.
The supported claim is that time matters, but not automatically. Longer residence can improve some outcomes, while barriers such as discrimination or non-recognition of qualifications can persist. The claim we cannot make is that time alone solves structural exclusion.
Myth: “Uncertainty is weakness.”
Reality: uncertainty is part of the evidence.
Uncertainty may come from sampling error, missing data, changing definitions, underreporting, or the fact that social processes have many causes. Researchers should not hide it. Journalists and advocates should not delete it to make a sentence sharper.
This does not mean every claim is equally uncertain. Some statements are well supported: source definitions matter; broad categories can hide important differences; official counts do not capture every experience; discrimination can affect outcomes; policy context shapes trajectories. Other statements are misleading: all migrants have the same impact, one group explains a complex social problem, or a single statistic proves success or failure.
The hardest category is “unknowable from the available data.” We may not know how many people avoided applying for a flat because they expected discrimination. We may not know how many people left a course because childcare failed, not motivation. We may not know whether a local service is excluding people unless it collects access data and asks those who did not return.
The practical standard is proportionate confidence. Say what the data source can show. Say what it cannot show. Say which categories are doing too much work. Then decide whether the public claim still deserves to be made.