Ingrid Vriens

Vriens-IngridTilburg School of Social and Behavioral Sciences
Tilburg University

Supervisors
Prof. J.K. Vermunt, Dr J.P.T.M. Gelissen & Dr G.B.D. Moors

On November 20th 2015, Ingrid Vriens defended her thesis entitled

Two of a kind?: Comparing ratings and rankings for measuring work values using latent class modeling

Kaft - Proefschrift Ingrid VriensSummary
The study of values is an important topic within the social sciences. After years of conducting research still uncertainty exists whether these values can be measured best by using a ranking approach or a rating approach.

In the ranking approach respondents are being asked to rank-order a number of alternatives based on importance of each of the alternatives relative to the other alternatives presented (for example: choose top 3 most important alternatives), while in the rating approach respondents are being asked to rate each of the alternatives on a predefined scale (like for example a 5-point Likert scale ranging from 1 “not important” to 5 “very important”).

Ideally results obtained by both methods should be similar, however there are method-specific features that may be biasing the results which have nothing to do with the content of the question (like the tendency in a rating scale to assign the same value to (almost) all alternatives).

The main finding in this dissertation is that, when controlling for method-specific features of each response format and by using a latent class modeling approach, the results obtained by the ranking and rating approach are more similar than previously assumed.

Project

Comparing rating and ranking procedures for the measurement of values in surveys
The study of values lies at the heart of the social sciences. Nonetheless, empirical social researchers have been involved in a long-standing discussion about the proper measurement of human value orientations, which revolves around the use of rating or ranking procedures. This project examines the appropriateness of both approaches in much-needed and novel ways, by : 1) directly considering the effects of response bias, 2) gathering and analysing data based on within-subjects survey experiments, which are from a Dutch nationality representative sample, and 3) making use of recent developments in statistical modelling of response styles and of rating and ranking data.

This project was financed by NWO.