Statistical Experimentation for Household Surveys

📖 The Scoop

The methodology and findings of the estimation of non-sampling error have wide applicability. This paper distinguished between enumerator error and respondent error and presents a model to measure the frequency of error occurrence. The model has limited applicability to measuring the extent of error. The data used derive from two case studies of Hong Kong. The major results include: (i) precision increases with a reduction in participation period matched by an increase in sample size; (ii) the use of multiple diaries within the household has no significant affect on response factors or level of expenditures reported; (iii) an inducement fee paid the household improves the response rate and does not significantly affect the level of expenditures reported; (iv) requesting income by source rather than as a total amount seemed to reduce underreporting; (v) one person households can be included; and (vi) difficulties exist in filling out daily expenditure records. This paper demonstrates the usefulness of undertaking such a methodological experiment as a way of improving survey data, and indicates the need for documenting such experiments thereby improving the efficient use of information resources.

Genre: Business & Economics / Economic Conditions (fancy, right?)