Tracking Polls

A tracking poll is a series of individual surveys repeated continuously over time to measure attitudinal and behavioral changes in a target population. While most commonly associated with election campaigns, tracking polls also are used for a variety of other research needs, from measurement of consumer sentiment to ad tracking in marketing research.

The term tracking often mistakenly is used to describe any trend data obtained over time. In fact it correctly applies only to continuous measurements in stand-alone samples. In all applications, the aim of a tracking poll is to produce an ongoing, rather than a one-time or episodic, assessment of evolving attitudes.

Tracking polls provide substantial flexibility in data analysis. The consistent methodology produces useful time trends. Tracking data can be segregated before and after an event of interest—a major policy address, a campaign gaffe, or an advertising launch—to assess that event's impact on attitudes or behavior. These same data can be aggregated to maximize sample sizes for greater analytical power. And tracking surveys can be reported in rolling averages, adding new waves of data while dropping old ones to smooth short-term or trendless variability or sampling noise.

The most publicized use of tracking polls is in election campaigns, particularly in the often-frenzied closing days of a contest when campaign advertising spikes, candidates voice their final appeals, voter interest peaks, and tentative preferences become final choices. Campaigns conduct their own private tracking polls to find their best prospects, target their message, and gauge their progress. The news media use tracking polls to understand and report the sentiments behind these crystallizing choices and to evaluate preferences among population groups, as well as to track those preferences themselves.

Election tracking surveys customarily are composed of a series of stand-alone, 1-night surveys, combined and reported in 3- or 4-day rolling averages. In well-designed election tracking polls, the limitations of 1-night sampling on dialing and callback regimens are mitigated by sample adjustments, such as a mix of new and previously dialed sample and the integration of scheduled callbacks into fieldwork protocols.

Election tracking polls have acquired a negative reputation in some quarters because of their reputed volatility. Thoughtful analysis by Robert Erikson and his colleagues, however, has established that volatility in pre-election polls, where it exists, is introduced by idiosyncratic likely voter modeling rather than by tracking poll methodology.

Humphrey Taylor, chairman of the Harris Poll, is credited with creating the first daily political tracking polls, a series of twice-daily, face-to-face surveys done for the Conservative Party in the last 4 weeks of the 1970 British general election: One survey tracked media exposure, while the other tracked issue priorities and party preferences on the issues. Taylor (personal communication, January 29, 2007) relates that his tracking data picked up “a collapse in confidence in the Labor government's ability to manage the economy following bad economic news three days before the election. And this was the reason why almost all the polls which stopped polling too soon failed to pick up a last-minute swing to the Conservatives.” The Tories won by two points.

...