Skip to main content icon/video/no-internet
icon backReturn to Entries

Total Survey Error (TSE)

Total survey error (TSE) is a term that is used to refer to all sources of bias (systematic error) and variance (random error) that may affect the validity (accuracy) of survey data. Total error in surveys can be conceptualized and categorized in many ways. One traditional approach is dividing total error into sources of sampling error and sources of nonsampling error. Another categorization is dividing it between coverage error, sampling error, nonresponse error, and measurement error. A more modern approach is to group various sources of error into the classes of representation and measurement. This entry provides a big picture perspective on all of the major types of error that occur in surveys and thus comprise total survey error.

Unfortunately, there is no such thing as a survey without error. Nevertheless, survey methodologists and survey practitioners aim for the most accurate surveys that can be conducted given the finite budget available to fund them. The quality of a survey statistic such as a mean, a percentage, or a correlation coefficient is assessed by multiple criteria: the timeliness of reporting, the relevance of the findings, the credibility of researchers and results, and the accuracy of the estimates—just to mention a few. Among those criteria the accuracy of the estimate is not necessarily the most important one. However, the accuracy is a dimension of the overall survey quality for which survey methodology offers a wide range of guidelines and instructions. Also, standard measures for the magnitude of the accuracy are available. The accuracy of a survey statistic is determined by its distance to or deviation from the true population parameter. If, for example, a survey aims to determine the average household income in a certain population, any deviation of the sample estimate from the true value—that is, what would have been determined if all members of the target population were asked their income and they all answered accurately—would decrease accuracy.

Representation and Measurement

There are two types of survey error that harm the accuracy of a survey estimate: random error and systematic error. Whereas random errors are assumed to cancel out each other—that is, negative deviations of the measurement from the true value are compensated by an “equal” amount of positive deviations—systematic errors shift the sample estimate systematically away from the true value; for example, because of certain question wording, respondents in a survey may tend to report a higher number of doctor visits than actually occurred in a given reference period. For linear estimates (such as means, percentages, and population totals), an increase in the random error leads to an increase in variance, whereas a rise in any systematic error results in an ascended bias of the survey estimate. The accuracy of a survey estimate is affected by either an increase of the bias or by a rise of the variance.

In a traditional view, the driving factors or sources of those survey errors are differentiated into two groups: sampling error and nonsampling error. Non-sampling error would then be further differentiated into coverage error, nonresponse error, and measurement error—some older textbooks mention processing error as well. However, a more modern theory-driven approach differentiates observational errors and nonobservational errors. While observational errors are related to the measurement of a particular variable for a particular sample unit, nonobservational errors occur when a net sample is established that is supposed to represent the target population. Following this path, Robert M. Groves and his colleagues have grouped the sources of error into two primary classes: representation and <>

...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles

Read next

More like this

Sage Recommends

We found other relevant content for you on other Sage platforms.

close

Have you created a personal profile? Login or create a profile so that you can save clips, playlists and searches

close
close
close Signed In to profile You are signed in as:
close
Logged in Institution Institution