Quality and Quantity 8 (1974) 297 9 Elsevier Scientific Publishing Company, Amsterdam - Printed in The Netherlands ABSTRACTS

ARE THERE STRUCTURAL

MODELS OF VOLUNTARISTIC

SOCIAL ACTION? HAYWARD R. ALKER, Jr.

Using a broad characterization of recent liberal, pluralistic, social science research interests defined in terms of Parsonian social theory, the paper reviews recent developments in social science statistical practices. These developments are seen as converging toward a cross-disciplinary synthesis of rigorous "structural" or "causal" modelling approaches. The major negative thesis of the paper is the frequent inadequacy in empirical practice of simply specified, statiscally manipulable structural models. They fail wholistically to handle emergence phenomena of primary interest to voluntaristic social theorists. A positive thesis is the possibility, in principle, of respecifying, and hence manipulating statistically, certain more complex simulational specifications of social action. Two subjects of social science research are reviewed as evidence for these theses: sequential prisoners dilemmas, and the United Nations collective security system. Earlier Markov learning models and path analyses with unmeasured variables were rejected in both cases by researchers now using simulation models with empirically more adequate results. The pedagogical implication is drawn that graduate training programs in social science departments should pay more attention to specificational alternatives, and relatively less to estimational efficiency.

CHANGE AND UNRELIABILITY OLOF PETERSSON

The article focuses on the robustness of turnover measures for panel data. The analysis is concentrated on the dichotomous case: two observations, and two response classes. Using a latent structural approach, two true latent classes are anlytically separated from two manifest response categories. Under the assumption of no true change in the latent classes, it is shown how the observed pattern of change is a function of a) the unreliability, and b) the variance of the measured attribute. Some examples demonstrate under which combinations of conditions s o m e measures of turnover differ most drastically from their true values. Some measures have often been used too uncritically in empirical panel analysis. It is shown how false conclusions about the true pattern of change can easily be made if the analyst neglects the impact of variance and unreliability.

297