Publication Type: Conference Paper/Unpublished Manuscript
Abstract: In this paper, the scientific productivity of the Finnish university system during 1999-2003 is analyzed first by disciplines and then by university units. The basic criterion for productivity is that the share of output should correspond to the share of input. Relying on unit-specific productivity coefficients, resource-related results are further compared between disciplines, with a special emphasis on the technical sciences. The intentions of Finnish ‘state-led technology policy’ are compared with the achievements of science policy in academic terms broken down by disciplinary differences in i) scientific productivity, ii) educational activities indispensable to scientific productivity and iii) the expansion of service and other activities insignificant to scientific productivity. The consequences of increasing involvement of external players to scientific actions, in the name of the ‘information society’ and ‘service universities’, are also considered in the broader context of a supply-driven higher education system built upon the traditional Humboldtian ideal: a multi-faculty university concentrating strictly on basic research and education based on it.

Publication Type: Conference Paper/Unpublished Manuscript
Abstract: The author attempted exit polling during advance poll voting and the general election in January 2006 to understand how and if party and candidate mobilization occurs. The polling was generally scuttled by Elections Canada and by other problems involved in trying to conduct exit polling remotely. The paper shares the problems that occurred and a few results reached by a very small N and offers suggestions to scholars who desire to exit poll in Canada in the future.

Publication Type: Paper/Poster Proposal
Abstract: As part of the ongoing development of the Consumer Assessment of Health Plans Survey (CAHPS®), researchers noted the need for a set of screening questions to identify persons with mobility impairment (PWMI). Although there is an established set of screeners to identify adults with chronic health conditions, previous research suggests that PWMI are a distinct subgroup of this population. Health plans in both the private and public sectors have an interest in learning more about the experiences of PWMI in order to improve the services they provide to this group. The first step toward achieving this goal was the development and testing of a series of eleven screening questions.

The screening questions were tested tested on a sample of 1124 adults enrolled in the Massachusetts Medicaid program (MassHealth). Sampled adults were sent a questionnaire and fact sheet that answered some commonly asked questions about the instrument. Reminder postcards were sent to everyone two weeks later, and a second mailing was sent to nonrespondents about two weeks after that. In accordance with the CAHPS® data collection protocol, professional interviewers then attempted to complete the questionnaire with nonrespondents by telephone. A total of 564 questionnaires were completed for a response rate of 51% (RR1, AAPOR 2004).

In order to validate and refine the new PWMI screening questions, medical claims data including ICD-9 codes were obtained for the survey sample.

Analyses of the data from both the questionnaire and medical claims allow us to address several research questions: Do the new screening questions appropriately identify PWMI? Is this group truly different from the group identified by the chronic condition screener? Can the 11-item series be shortened to reduce respondent burden without degrading its effectiveness? Can the same PWMI be identified by using claims data alone, negating the need for screening questions?

Publication Type: Conference Paper/Unpublished Manuscript
Review Method: Peer Reviewed
Abstract: Research panels have some advantages for longitudinal research in that the costs of the initial recruitment can be spread out over multiple studies or waves of data collection. However, all studies conducted using research panels, as distinct from those based on cross-sectional samples, are affected by the loss of research panelists prior to sample selection. A research panel that contains a mix of recent recruits and more tenured research subjects provides some insurance against nonresponse bias resulting from panel losses; however, the question remains how much the unavailability of the lost panelists affects survey estimates.

This paper examines the effects of panel attrition on substantive survey estimates regarding political attitudes and behavior, health status, lifestyle, and other areas, in internal surveys conducted by Knowledge Networks. We compared survey estimates based on all collected responses to survey estimates based on a subset of responses corresponding to non-attrited cases (i.e., currently active panel members).

Our analyses of the survey data suggest that substantive survey results are slightly affected by panel attrition after application of poststratification weighting for several demographic characteristics. We compared the weighted survey estimates for active panel members versus all active and attrited panel members for 30 substantive survey variables (n=15,000 to 48,000 responses) from our health and political profile surveys. Large differences in these estimates would indicate that lost panel members reported different responses than active panelists. The average absolute difference across the variables is small (0.009 or less than one percentage point), indicating that the loss of attrited panel members is having a relatively small impact on data quality. An additional 300 variables were also examined across a broad range of topics, and again the average absolute differences are small (between 0.003 and 0.015). Examples from specific studies will also be discussed in the paper.

Publication Type: Conference Paper/Unpublished Manuscript
Review Method: Peer Reviewed
Abstract: Longitudinal studies offer the opportunity to examine trends over time both in terms of participation and attitudes. However, the loss of participants from one wave to another can result in bias of survey results. The Study of HIV and AIDS Stigma, conducted for the Centers for Disease Control (CDC), is a longitudinal study of a national, RDD selected Knowledge Networks panel investigating knowledge of HIV and AIDS, as well as attitudes and beliefs surrounding this issue. The survey was initially conducted in summer 2000 with a follow-up in spring 2003. Comparisons are made of panel participants and non-participants between the 2 waves, as well as changes in attitudes. Survey non-participants at time 2 include non-respondents who are still on the panel, and non-respondents from attrition (who have chosen to withdraw from the panel). In addition, we investigate changes in attitudes over time and discuss the implications of differing panel participation on those changing attitudes and how this may result in biased data. Early analysis indicates that participants with more negative attitudes in wave 1 were less likely to participate in wave 2 resulting in potential bias to the results.