Beschreibung

"Heaping is a response error typical to retrospectively collected labor force status data. Respondents round the spell length when duration data is collected by episode-based questionnaires. Calendar-based questionnaires instead lead to abnormal concentrations of the start and/or end of spells at specific calendar months. By Monte-Carlo simulation we quantify the effects of heaping on parameter estimates of Weibull and log-logistic models that imply a median duration that is realistic for unemployment duration data. When there is upward and downward heaping and the measurement error is approximately zero-mean, the parameter estimates of the standard duration models are very robust. As far as we consider only upward heaping, we find that a large amount of heaping leads to substantial inconsistencies and the biases are approximately linear in the amount of heaping." (Author's abstract, IAB-Doku) ((en))