A tracking poll is a repeated political survey conducted using substantially similar questions and methodology over multiple points in time so that changes in public opinion can be monitored. Instead of producing only one snapshot, a tracking poll creates a sequence of measurements that can be examined for movement, stability and short term fluctuations. Tracking can be used for voting intention, approval, favourability, issue preferences or other political measures. The central concept is temporal comparison.
If a question produces one result in one period and another result later, researchers can investigate whether the difference appears consistent with genuine movement or could instead reflect ordinary sampling variation and methodological noise. The value of a tracking poll therefore lies in observing change rather than merely reporting a single percentage.
Tracking polls require methodological consistency because comparisons across time become harder to interpret if the underlying survey design changes substantially. A pollster may keep the target population, question wording, response options, sampling approach and weighting framework broadly consistent while collecting new responses at regular intervals. Some tracking designs use overlapping samples or rolling samples, while others conduct repeated independent samples. A rolling design may continuously replace older interviews with newer interviews, creating a moving window of observations. The exact design matters because two results collected from completely separate samples are not equivalent to two results derived from heavily overlapping respondents. Transparency about the fieldwork period and sample construction is therefore necessary for interpreting apparent changes.
One important reason for tracking polling is that political attitudes can change in response to events, campaigns, debates, economic developments, political decisions, scandals, crises, announcements and other information. A single poll can show the level of support at one particular point but cannot by itself establish whether support is increasing, decreasing or stable. Repeated observations make it possible to examine patterns across time. However, not every movement between two polling results represents genuine public opinion change. Sampling variation alone can cause results to move upward or downward, and changes in question wording, sample composition, weighting or data collection mode can also produce differences. AAPOR specifically recommends considering the timing of fieldwork and methodological differences when comparing polls.
Tracking polls can be constructed using different sampling arrangements. A repeated cross sectional design surveys different people from the same target population at different times, allowing population level change to be estimated without requiring the same individuals to participate repeatedly. A panel design follows some or all of the same respondents over time, allowing researchers to examine individual level changes in opinion. These designs answer different questions.
Repeated cross sectional polling is primarily suited to detecting changes in the distribution of opinion across a population, while panel research can investigate how particular people change their views. Panel studies introduce additional issues such as attrition, conditioning effects and repeated participation, so they are not simply interchangeable with ordinary tracking surveys.
A rolling tracking poll can provide a smoother representation of short term movement because each published estimate may combine responses collected across a defined time window. For example, a pollster might continually collect interviews and report the most recent several days of responses as a moving estimate. This approach increases the amount of data contributing to each published result but means that consecutive estimates may share respondents or fieldwork periods. As a result, adjacent observations are not necessarily statistically independent. Apparent movement can therefore be smaller or larger in significance than a simple comparison of two percentages might suggest. Analysts need information about the rolling window, overlap and sampling design to interpret the trend correctly.
Tracking polling is particularly useful for distinguishing persistent movement from isolated results, although it cannot completely eliminate uncertainty. If several consecutive measurements move in the same direction using comparable methodology, the evidence for a sustained change may be stronger than a single change between two unrelated polls. Nevertheless, researchers should consider the size of the change relative to sampling uncertainty and other sources of error. AAPOR notes that poll results can differ because of sampling error, respondent recruitment, turnout modelling and question wording, and it recommends examining trends across multiple polls rather than treating one poll as definitive.
Tracking polls can be used for many measures besides candidate preference. A political organisation might track approval of a leader, favourability towards a political party, concern about an issue, satisfaction with government performance or support for a policy. Because the same measure is collected repeatedly, researchers can create a time series showing how the measured attitude changes. The interpretation still depends on the exact question. A movement in approval is not necessarily a movement in voting intention, and a change in an issue's importance is not necessarily a change in policy support. Tracking provides continuity of measurement, but continuity does not make different political concepts equivalent.
The principal limitations of tracking polls arise when apparent trends are mistaken for certainty or when methodological changes are ignored. A poll can track the wrong population, repeatedly reproduce a biased sampling process, or consistently measure a question imperfectly. If the composition of respondents changes, weighting may become important, but weighting itself can introduce modelling assumptions. If the question wording changes, a break in the time series may occur even if public opinion has not changed. Tracking polls therefore provide evidence about change under a particular measurement system, not an infallible record of the public mind. Their proper interpretation requires attention to fieldwork dates, sample design, weighting, question wording, response rates and the statistical uncertainty surrounding each estimate.
