What a “commute” is in the data
A commute sounds like a clear thing: the journey between home and work. Surveys, though, must turn that into a question with a fixed set of answers, and decisions made at that stage shape every figure that follows. Does the question refer to the main job or all jobs? To a normal day, or to one specific day? To the longest part of the trip, or the whole of it? Different sources answer these differently, which is one reason commuting statistics from two organisations rarely match exactly.
Where the data comes from
Censuses. Many national censuses ask where people work and how they usually travel there. Because coverage is close to complete, census answers can be broken down by small areas and used to map flows between neighbourhoods and towns. The questions are brief and fixed, and the answer refers to a particular reference day or week, so the picture can be dated and coarse.
Household and travel surveys. Dedicated surveys ask a sample of households about their travel in more detail, sometimes with a travel diary covering all trips on a set day. These capture more than commuting, including shopping and leisure trips, and can record the stages of a journey. Because they use samples, results carry sampling uncertainty and are typically less reliable for small places. How well they reflect the whole population depends on the sampling frames used and on who responds.
Labour force and other regular surveys. Some ongoing surveys include a travel-to-work question as a side topic. These are good for tracking change but may use a shorter question than a travel survey.
Counts and sensors. Traffic counters, ticket validations and pedestrian or cycle counters record movement directly, without asking anyone. They show how many pass a point, not who they are or where they are going. Transit tap data covers only passengers who use that system.
Mobile location data. Anonymised phone signals can suggest patterns of movement at scale. They depend on who carries a device and which apps share location, and they infer rather than record the purpose of a trip.
The problem of one reported mode
Most census-style questions ask for a single “usual” or “main” mode of travel: car, train, bus, bicycle, on foot and so on. This is convenient but loses a great deal. Real journeys are often mixed. Someone might cycle to a station, take a train, then walk to the office. Another person might drive to a car park, then take a tram. A single answer must choose one of these, and the rule used matters.
Common approaches include recording the mode used for the longest distance, the longest time, or the one the respondent considers main. Each rule changes the counts. A short walk at either end disappears under a long train ride. A cycle leg to a station might never show up in the headline share of cycling, even though many journeys involve it. Where a source’s headline says a small share of commuters walk or cycle, that can understate how many include a walking or cycling stage.
The same logic applies to the idea of “usual”. People whose mode varies by day, weather or season may pick the one they use most, or the one that comes to mind first. Someone who drives three days and takes the train two may be recorded as a car commuter.
Variation across the week and the year
A usual commute is a simplification. Many people do not travel to the same place on the same days. Shift workers, part-timers, people with several jobs and those who work partly from home have irregular patterns. The growth of hybrid work makes this more visible: a person may commute two or three days a week, and a question about “your usual journey” may capture a day that is not typical of their week.
This also complicates comparisons over time. If the wording of the question stays the same but working patterns change, the result can shift for reasons that are not about transport. Someone who works at home most days may answer a question about their usual mode by describing a trip they rarely make, or may be placed in a separate “works at home” category, depending on how the form is designed. Question wording thus affects not only the answers but also who is counted as a commuter at all.
Defining the denominator
Commuting shares are only meaningful with a clear base. A figure for “share of workers who drive” differs from “share of all residents”, and from “share of trips”. It may include or exclude people with no fixed workplace, such as delivery drivers or construction workers who move between sites. It may cover only those in employment, leaving out students and others making regular journeys.
Place matters too. Statistics can be organised by where people live or by where they work. A city centre might look very different when counting workers who come in than when counting residents who leave. Averages across a whole region can hide very different experiences in a dense centre and a distant suburb, and treating a regional average as if it described each individual is the ecological fallacy at work.
What commuting figures can and cannot show
Commuting data is good for showing broad patterns: which modes are used most for work trips, how flows differ between areas, and whether shares shift over several years under consistent definitions. It is weaker for:
- Stages and combinations of modes, when only one is recorded.
- Non-work trips, which make up a large part of travel and may follow different patterns.
- Day-to-day variation, which a single “usual” answer smooths over.
- Reasons for choices, since the data records what, not why. Cost, availability, distance, employer policy and personal preference all interact.
When a news item links a change in commuting to a single cause, such as a new transit line or a policy change, it is making a causal claim that survey shares alone seldom establish. Other things usually change at the same time.
Reading a commuting statistic
Useful questions include: Is this a main mode or all modes? A usual day or a specific one? Residents or workers? Which year’s definitions are being used? A figure without these answers can be technically correct and still easy to misread.