Estimating recall bias without gold standards: job tenure in Colombia

Abstract

This study proposes an alternative method for estimating the incidence and severity of recall bias that does not require a frequently inaccurate ‘gold standard’. Analysing job tenure from Colombian panel data, we concluded that contextual factors are more relevant than individual socio-economic characteristics in shaping recall bias outcomes.

Notes

¹Excluding responses that exceed two SDs imply excluding observations that lead to biases of more than ± 132 months, or 11 years. Among the reasons for such implausible answers we found that some answers for 2007 appeared to be expressed in years, whereas in months for 2008. Others were simply nonsensical as the tenure in 2007 was reported to exceed that in 2008 by, at times, dozens of months.