Measurement errors Sample Clauses

Measurement errors. The turnover figures extracted from VAT returns are subject to large measurement errors. And even though ONS apply three automatic rules for cleaning VAT returns (see Xxxxxxxx and Allcoat (2016)), they remain very noisy. Some of this noise could arise from the difficulty in linking VAT registration units to the business register used by ONS. If a VAT return relates to turnover arising from a number of distinct business activities, the turnover needs to be attributed between those activities. An issue arises as well when more than one VAT returns relate to a single business unit. ONS attribute VAT turnover to units from the business register using employment size. But some firms have complicated business patterns and carry out activities in several industries, making the attribution of turnover difficult. This problem of apportionment is more prevalent in large firms. Overall the sources of measurement errors affecting the VAT returns are not well identified and understood. And the extremely large size of the VAT sample means that it is not possible to clean individual returns manually. As a result the VAT data analysed contain substantial measurement errors which complicate their interpolation. In this context least-squares techniques are not a viable means of interpolation. Consequently, an unobserved components model is developed in chapter 3 to accommodate the noise embedded in the figures as well as their overlapping nature.‌