A family sits down after dinner and turns on the television. Market researchers record what they are watching and repeat this for thousands of families. But is the information they collect as useful as it could be?

Households are often made up of several individuals. Are they all watching the television? Who has sway over what is watched? These questions are important to both programmers and advertisers.

Ying Zhao of HKUST and his co-authors Sha Yang, Yi Zhao and Tülin Erdem try to come up with answers through a model that captures the behavioral interaction between members of the household over time, when they are presented with multiple choices. The model is tested in the context of television viewing.

"Some family members may be more likely to watch television with other members because they are likely to receive utility from spending time together with their spouses, children or parents. It is this behavioral interaction among family members that we study," they say.

The model is tested on 187 families with at least one child aged two to 18, over a four-week period in 2002. Five types of programs are measured - news, movies, kids, talk and sports - and the families are contacted every 15 minutes from 6.30pm-10.15pm to see if they are watching television, what they are watching and who they are watching it with. The information gathered reveals their individual preferences and joint interactions over time.

Tellingly, the "household preference" does not reflect individual preferences. The household figures are based on overall frequency of viewing. In the sample for news programs, the figure is 41.38 per cent overall. But when that is broken down at the individual level, the figure for fathers is 40.41 per cent, for mothers 46.05 per cent and for children 41.7 per cent. For sports programs the overall frequency is 17.04 per cent but for fathers it is 25.03 per cent, for mothers 13.17 per cent and for children 14.25 per cent. Such variations are seen for all five types of television programs.

"These differences between the family viewing patters and individual members' viewing patterns suggest that the household-level analysis does not capture each household member's viewing behavior well," the authors say.

They break this down in terms of viewing partners and show that father-mother behavioral interactions are strongest followed by mother-child and father-child. Certain family characteristics also affect these interactions. For instance, higher income increases father-child interaction on kids' programs, although this is smaller in non-white families, families with Internet connections and those with older fathers. Mother-child interactions are higher for non-white families when it comes to news programs, for families with higher education when it comes to movies, and for families living in larger cities when it comes to sports.

Such factors as income, number of children, age and family structure (whether single-parent or not) are also shown to affect individual preferences.

"Our model estimates can help companies better understand the influence of intrinsic and extrinsic preferences on individuals' overall viewing preferences and more importantly, the influence of characteristics of television programs and individual/family type on these intrinsic and extrinsic preferences," the authors say.

Consideration is also given to who has the power to determine what is watched. This is useful for programmers because they can promote their programs to these people. The authors find much variation. For example, mothers in non-white families have a higher power in the case of news programs as compared to mothers in white families, and fathers in non-white families have more power in the case of sports programs.

The authors also look at how decisions are made and find a variety of tactics are used. Families may be guided by the average of family member preferences, the person with the lowest preference or the person with the highest preference.

Overall, the findings add to marketers knowledge, thus enabling them to target their communications more accurately and potentially influence the family purchase and consumption behavior.