Estimation of Dynamic Linear Models in Short Panels with Ordinal Observation of the Endogenous Variables

Publication type

Conference Paper


Italian Congress of Econometrics and Empirical Economics


Publication date

January 24, 2005


This research is concerned with the use of subjective measures of financial well-being as indicators to be used in welfare analysis as an alternative or supplement to income variables. The BHPS contains questions asking respondents to summarise how well they are managing financially, using an ordinal scale. This is asked separately about the current period and about change since the previous year. The paper proposes and implements a new statistical approach to modelling these responses when there is the possibility of slow dynamic adjustment of perceptions. This approach uses a latent autoregressive or autoregressive-moving average structure to model the underlying continuous perceptions, with these latent variables converted into ordinal form via the usual ordered probit mechanism. We use simulated maximum likelihood estimators, with the GHK simulator used to evaluate the required high-dimensional rectangle probabilities. We demonstrate that this model fits the BHPS considerably better than the more common Heckman state-dependence model. We test the hypothesis that responses to the questions on current financial status and recent change are mutually consistent. The results suggest that perceptions are subject to significant adjustment delays and that there may be persistent inconsistencies between responses to the current and retrospective questions.

Related Publications



Latest findings, new research

Publications search

Search all research by subject and author


Researchers discuss their findings and what they mean for society


Background and context, methods and data, aims and outputs


Conferences, seminars and workshops

Survey methodology

Specialist research, practice and study

Taking the long view

ISER's annual report


Key research themes and areas of interest