Where does tracking collide?

In control engineering, tracking is the task of making a system's output follow a commanded reference or a moving target over time, so that the difference between the two stays small. Hazen's 1934 paper Theory of servo-mechanisms set out the mathematics of a servomechanism, a device that detects the gap between a desired position and the actual one and drives a motor to close it, which is the origin of treating following a reference as a formal design goal. Later target-tracking systems, such as radar-directed gun controllers, extended the idea to a reference that moves unpredictably, and modern control texts separate tracking, following a changing setpoint, from regulation, holding a fixed one. Here the referent is a machine chasing a signal, judged by how closely the output matches the command, not the sorting of pupils, the copying of a market index, or a child staying in a growth channel.[1]

Unit

One reference-following task, scaled by the error between commanded and actual output

Recovery means

After a disturbance the feedback loop drives the tracking error back toward zero

Measured by

Quantified from the error signal over time, for example steady-state error, settling time, or integrated squared error

In finance, tracking is the practice of building a portfolio that reproduces the return of a chosen market index as closely as possible, and tracking error is how far the portfolio's return drifts from that index. Renshaw and Feldstein's 1960 article The Case for an Unmanaged Investment Company first argued for a fund that simply holds a market average rather than trying to beat it, the idea behind index investing. Roll's 1992 A Mean/Variance Analysis of Tracking Error gave the modern formal treatment, defining tracking error as the standard deviation of the difference between portfolio and benchmark returns and showing how a manager trades it off against expected return. Here the referent is a portfolio shadowing an index, measured in percentage points of return deviation, not a controller following a signal, students placed in curricular streams, or a growth measurement holding its rank.[2][3]

Unit

One benchmark-replicating portfolio, scaled by its tracking error against the index

Recovery means

Not applicable

Measured by

Tracking error computed as the standard deviation of periodic return differences between portfolio and benchmark

In the sociology of education, tracking is the sorting of students into separate instructional groups or curricular streams, such as college-preparatory, general, and vocational, usually by measured ability or past achievement. Rosenbaum's 1976 book Making Inequality: The Hidden Curriculum of High School Tracking documented how these placements, once made, tend to lock students onto divergent paths. Oakes's 1985 Keeping Track: How Schools Structure Inequality became the seminal study, showing that track placement shapes what content students are taught and how opportunity is distributed among them. Here the referent is an administrative grouping of children within a school, described by placement rules and their consequences, not a machine following a reference, a fund copying an index, or a body maintaining a growth percentile.[4][5]

Unit

One student's placement into a track or ability group within a school

Recovery means

Not applicable

Measured by

Observed through placement records, course-taking patterns, and the share of students who move between tracks

In developmental medicine and epidemiology, tracking is the tendency of a person's measurement, such as height, weight, blood pressure, or blood lipids, to stay in the same relative rank or percentile channel as they grow older. Clarke and colleagues' 1978 study Tracking of blood lipids and blood pressures in school age children, part of the Muscatine study, gave the term this longitudinal meaning by showing that children tended to keep their rank on repeated measurement across years. Foulkes and Davis's 1981 An Index of Tracking for Longitudinal Data then supplied a formal statistic for how strongly a measurement holds its position over time. Here the referent is a biological trait keeping its place in a population distribution, measured by rank agreement across repeated observations, not a controller chasing a target, a portfolio copying an index, or a school sorting pupils.[6][7]

Unit

One individual's percentile position for a trait, followed across repeated measurements

Recovery means

Not applicable

Measured by

Assessed by the correlation of a person's rank across time points, or a tracking index such as that of Foulkes and Davis