Panel

Advantages and Disadvantages

There are clear advantages to using a panel rather than using a series of independent samples in studying a target population. Some of these include the following:

• 1.
  A panel provides the details on the nature of change. For example, suppose that from one independent sample to another the prevalence of a disease changes from 5% to 15%. We know the [Page 565]simple change of 10 percentage points, but do not know whether the incidence of new cases is 0% or 10% or something in between 0% and 10%. Using a panel, the percentage of new cases is easily obtained.
• 2.
  It reduces the variability for estimates of change at different points in time. For example, if P1 and P2 are the estimated proportion of unemployed people at Time 1 and Time 2, respectively, then the estimate of change is P2 − P1 When using a panel, the variance of P2 − P1 is reduced relative to the one for two independent samples.
• 3.
  It often reduces the observational errors by interviewers or respondents. For example, as the surveys are repeated, interviewers have a better experience of administering the interviews and respondents have a better understanding of the questionnaire.
• 4.
  It gives information on the dynamic behavior of respondents over time. For example, it is possible to explain the fact that the same people experiencing an event in the past, say unemployment, tend to experience it in the future.

However, there also are disadvantages to using a panel:

• 1.
  Analysis of a panel data is more complicated due to differential unit and item nonresponse and wave non-response, as well as the complexity of sample design. Wave nonresponse occurs when one or more waves of panel data are missing for a sample unit that has responded for at least one wave. Weighting, imputation, or a combination of weighting and imputation can be employed to compensate for missing wave data.
• 2.
  Measuring changes in individuals from one year to another year may be unreliable because reluctant respondents may give poor answers to repeated interviews, or respondents may refuse to be interviewed several times due to panel fatigue, resulting in higher nonresponse over time.
• 3.
  Respondents' answers to questions in previous waves may affect their answers in subsequent waves; this is termed panel conditioning.
• 4.
  It is difficult to keep the panel representative during a long period of time because the target population can change over time.
• 5.
  It may be too expensive to locate all respondents a year later or after a certain period of time, due to travel costs and the obstacles to following and finding some respondents who have moved without any new contact information.
• 6.
  It can be difficult to identify the same sample units over time. For example, identification of the same family units can be complicated when the family composition is changed by births, deaths, marriages, divorces, and so on.

Rotating Designs

Even if a panel is a “bad” sample, that is, if it does not well represent a given population over time, the organization carrying out the panel survey may have to continue to maintain that panel. But there is a solution to such a problem of the panel. It is a rotating panel design, in which a part of the panel sample is replaced at each subsequent point in time. This design is intermediate (i.e. a hybrid) between a panel sample and independent samples. As the simplest example, one may choose a panel design involving overlaps of half of the sample elements, as shown by AB, BC, CD, DE, EF, and so on at different points of time. In this example, the B panel sample appears in the first and second waves of data collection; the C panel sample appears in the second and third waves, and so on. Such rotating designs not only reduce respondent burden but also provide an opportunity to refresh the sample at each point of time with cases that better reflect the current makeup of the target population.

...