VAT part 2 - additional indicators, bias and anomalies

VAT part 2

This project covers further work that the Campus does on the VAT (expenditure and turnover data). The Project will:

  • explore the use of VAT returns as an early economic indicators (e.g. expenditure, births and deaths, number and timing of returns, modelling economic statistics such as profits etc) – this may include standalone indicators, inputs into improving national economic statistics, and insights into quality assurance of national economic statistics
  • explore bias and quality, for example within different reporting periods, assumptions about reporting level (enterprise, group etc.), types of return, editing etc.
  • develop methods for anomaly detection in the VAT data which can:
    • be used for quality assurance and economic analysis
    • be applied to other large administrative data sets *dynamic systems, network analysis, agent based models
  • can we model intermediate consumption form VAT TO & EXP, and subtract to get an estimate for cap ex?

Suggested priorities for discussion 16 Nov, for stage 1 of this, to happen over the next 3 months

All of these for the reference quarter, but could compare with later revised values. Use raw data - caveat for regional, SIC, employment breakdowns

  • update vindicator and publish article & data ASAP
  • births & deaths - number of each, growth rate of time series, difference (B - D)
  • number of returns in the reference quarter, and by (cumulative) month
  • can explore by SIC, region and employment
  • expenditure - as for turnover
  • clustering? including all variables, maybe drop TO / exp? does this tell us anything, and does the mix of reporting types change? -> VAT reporting behaviour

And some of the areas that could be explored include:

  • apportionment to regions - checking summing to totals, better methodology
  • aggregation to quarterly time series, including exploring biases within different reporting periods
  • biases & relationships - reporting unit, enterprise, enterprise group etc. - errors in assumptions of reporting level?
  • match < 100% QOPS – why, too small to match VAT? births & deaths? timing where VAT not reported?
  • HMRC VAT registrations - daily delivery
  • other VAT info - change to staggers, type of return e.g. rebate, etc - clustering, growth / decline, revisions – how can all of these be used to better understand what’s going on in the economy, or as early indicators that something new is happening
  • QA – identifying real outliers / anomalies, identifying anomalous behaviour in groups with certain characteristics?
  • Can we improve survey aggregation? Particularly for volatile / badly behaved indicators like capital investment?
  • Can this be generalised to other large admin datasets (e.g. PAYE)?

team members

  • Louisa Nolan
  • Jonathan Gillard (Reader in Statistics, Cardiff Uni)
  • Emily O’Riordan (PhD student, anomaly detection, Cardiff Uni)
  • Luke Shaw

stakeholders

  • Economic stats: Rob Kent-Smith, Andrew Sutton, Richard Heys, James Scruton, Rob Doody etc.,
  • Economic Experts Working Group
  • economics users
  • NSIs - speak to Mark Stephens
  • Martin Weale
  • Duncam Elliot, methodology tiime series

Updates

2018-11-19T16:52:54Z Project is resourced jointly with Data scientists and Economists.

Notes

This page has been automatically generated. Click here to download this project description as a pdf or click here to download as a word document.


Updated