Abstract
By measuring response errors in reports of savings account and debt balances under near-ideal (and very special) conditions, this article shows what is the maximum degree of accuracy we may expect to attain in sample surveys of a financial character. Under the conditions of this study none of the response errors observed were sufficiently serious to invalidate the data for any likely statistical use. The article investigates the influence on response errors of the following factors: record consultation, “rounding,” size of balances, the number and nature of transactions in the account balance, the length of the recall period.

This publication has 4 references indexed in Scilit: