The effect of coding error on time use surveys estimates

Download

PDF - Publishers print Restricted to Registered users only Download (131Kb)

Description/Abstract

This article presents the results of a coder reliability study conducted as part of the 2000 UK Time Use Survey. Five coders coded the same 40 diaries in which respondents had recorded, in their own words, their activities for every ten minutes over the course of a particular day. Coding was done via a computerised coding system, which enabled coders to view scanned digital images of diaries and access an online coding frame. In addition to an estimate of net aggregate coder reliability, proportion of agreement coefficients are presented for each of the ten main activity codes at the highest level of the coding hierarchy. Reliabilities are also calculated for individual coders. Intra-class correlation coefficients (Rho) are then estimated and these are combined with the reliability estimates to produce a variance inflation factor for each of the ten higher order main activity codes (). Some illustrative examples are provided to demonstrate the true standard errors of survey estimates once this coder error has been accounted for.