What's the issue?
Ideally, to understand how children use the internet at home, one would interview both children and parents, so as to triangulate the two data sources, to permit parents to provide a check on responses from young children, and to permit children to report on their experiences themselves, especially since parents may not be aware of the range of their activities and perceptions. However, this is complicated in terms of both recruitment and data analysis, and thus it is a relatively expensive approach to research. Researchers are therefore often left with having to decide which one to interview when their resources are limited.
Rules of thumb are to include both children and parents (or teachers) as respondents wherever possible.
One cost-efficient route to combining data sources is to ask just a few, key questions of parents when recruiting children.
If both can be included, children should be reassured that parents will not see their responses (cf FAQ on ethics).
In reporting, care must be taken when assuming that one set of responses are more 'correct' than the other - probably, it is safest to regard the discrepancy as indicating the upper and lower bounds for a response.
Note that, as a rule, children tend to report higher estimates of internet use and risk, and lower estimates of parental mediation and internet-related anxieties, compared with parents.
Pitfalls to avoid
If only parents/adults are interviewed, care must be taken in interpreting their claim if they relate to phenomena to which their access may be limited (e.g. accounts of what children do in their bedroom, in private, on their mobile or at school).
Only interviewing children, as in the Mediappro (2006) project, has other disadvantages: most notably, it is difficult to get reliable information on socio-economic status (whether parental income, education or some combination thereof), and so findings regarding inequalities or exclusion cannot be obtained. One solution is to sample schools in more and less advantaged neighbourhoods Hence the Mediappro project, which surveyed 7393 12-18 year olds in nine countries, used a stratified sample of schools in which researchers conduct ed a pen-and-paper survey; see http://www.mediappro.org/.
A researcher's experience
The SAFT surveys interviewed both children and parents, using the same questions for each. Where children and parents give fairly similar answers (e.g. 31% of children and 21% of parents say the child does instant messaging), the 'truth' may be taken to lie in between. But where answers are different (e.g. 56% of children but only 8% of parents say the child downloads music), it is clear that relying on parents to provide reliable information about children is insufficient and misleading. Furthermore, significant findings emerge precisely from these discrepancies. For example, since 64% of children say their parent never sits with them when they go online, while only 11% of parents say they never sit with their child, one can conclude both that children may be 'saving face' by underreporting how often a parent sits with them, but also that parents are both relatively ignorant of their children's actual use and overconfident of their own safety practices.
Example of weak practice
Less useful, by contrast, is the reliance in the Eurobarometer survey on adults reporting about children Although this survey has provided much useful information regarding children's and parents' internet use across Europe, it is significant that survey respondents were adults over 15 years old who were responsible for, or caretakers of, a child under 17 years old. Thus, not only does this survey of children's internet use rely on reporting by adults but these adults may not be the child's parent (but could be a child-care employee or older sibling, for example).
Example of good practice
The recruitment strategy used by the Youth Internet Safety Survey (Shade, 2002-2005) in the USA efficiently obtained two sources of data (asking a few questions of parents when recruiting children), got informed consent from both parents and children, and established an appropriate context for a sensitive interview, in a single telephone call as follows:
"When contacting a household, interviewers from a national survey research firm screened for regular use of the internet by a youth in the target age group. Interviewers then asked to speak with the parent who knew the most about the youth's internet use, conducted a short interview assessing household rules and parental concerns about internet use, and gathered demographic characteristics. The interviewer requested permission from the parent to speak with the youth. Parents were assured of the confidentiality of the interview and were informed that the interview would include questions about "sexual material your child may have seen." Upon achieving parental consent, interviewers described the study to the youth and obtained his or her oral consent. Youth interviews, which lasted about half an hour, were scheduled at the youth's convenience and arranged for times when he/she could talk freely."
(Page 3011 of Mitchell, K. J., Finkelhor, D. & Wolak, J. (2001). Risk factors for and impact of online sexual solicitation of youth. Journal of the American Medical Association (JAMA), 285(23), 3011-3014.)