Size limits for TESS experiments

The shorter an experiment, the more respondents on which it can be constructed. For each unit of experiment length, the maximum number of available respondents (subject to guidelines below) is as follows:

Calculating survey length

• Each question counts as 1 unit of length.
  
• If a survey contains a short paragraph of text (such as a vignette) in addition to the question(s), this counts as 1 unit.
• A longer paragraph of text (as in a multi-paragraph vignette) counts as 2 units
• Each picture included counts as 1 unit (the picture presented to respondents may differ for different experimental conditions).
• Playing a video clip of up to 30 seconds counts as 1 unit plus the programming surcharge. Up to four alternative clips may be included for a programming surcharge of 1 unit. (For example, consider an experiment in which each respondent will view 1 randomly-assigned clip from a set of 4 alternatives. Playing the clip would count 1 unit and programming would count 1 unit, for 2 total. Now consider an experiment in which each respondent will view 2 clips, each selected from 2 alternatives. Playing the clips would count 2 units and programming would count 1 unit, for 3 total.)
• Playing an audio clip of up to 30 seconds counts as 1 unit plus the programming surcharge. Up to two alternative clips may be included for a programming surcharge of 1 unit.
• In addition to the standard variables included with every TESS experiment, up to 3 AmeriSpeak profile variables may be included (see links for contents of the public affairs, health, and financial profiles, as well as the core adult profile which includes additional demographic variables to those included in the TESS standard delivery). More than 3 profile items can be obtained at a cost of one unit of survey length for each 6 items requested.

Calculating number of respondents

• The above guidelines presume that the experiment will be conducted without any subsampling. TESS experiments can be conducted on particular populations. However, subpopulations will reduce the size of the maximum sample from 10% to 50%. As an illustrative example, proposing to conduct the experiment only among those respondents who presently have children living at home would reduce the maximum by ~10%, while proposing to conduct the experiment only among those respondents who are Catholic and who are self-identified liberals will reduce the maximum by ~50%.
• The above guidelines presume the experiment that the programming demands of the experiment are not unusual for TESS studies. In cases of unusually high programming complexity, the maximum number of available respondents may be reduced by 10% to cover extra programming costs.

What if my experiment needs to be larger than what these limits allow?

• Experiments larger than these limits can be conducted if the investigator has funding from another source to cover the additional costs.

Additional details

• Short experiments: Studies less than 5 units in length may need to wait and be bundled with other TESS experiments for purposes of cost-efficient fielding. The maximum N for short experiments with subsampling will likely be the N implied by running the experiment on the full population.