Thursday, February 09, 2006
Wetting our pants?
Now here's an interesting comment on a previous post - covering the online self assessment service - from "homer" (how is bart?) ... I'm pasting it in here in its entirety (I've found it difficult to have a debate in the comments section - all that unfolding and stuff, so it seems to make sense to bring comments up to the top every so often to help keep track of the debate). "Well before we starting wetting our pants at what could be the first evidence of a working government website, let's have a quick look at how the FT report the same story today ... ... The income tax self-assessment system is riddled with errors, with at least a quarter of PAYE tax codes incorrectly calculated by the Revenue & Customs, a Commons inquiry will warn today. An estimated one in three income tax returns are incorrect, costing the state an "enormous" £2.8bn a year in lost tax payments, says a report by the public accounts select committee into the Revenue's performance. The report attacks the Revenue's own lack of accuracy and urges it to "accept more responsibility for its mistakes". Almost 500,000 returns - 5 per cent of the total - were processed wrongly in the last tax year. Never mind, a small detail. Want to buy some consultancy on xml, search engines, passport, ... ... ?" It's a good place to start a debate. Leaving aside that I was talking about the apparent success of the online service and that this is a report on a problem with the self assessment process overall, here are my thoughts: - Let's start with breaking the taxation process down into its component parts: - Legislation (tax law) - Data requirements (defined by the law, provided by the citizen or their employer) - Form design (defined by HMRC in response to the data needed) - Form completion (up to the citizen/employer or their accountant) - Form submission (post or online, choice of the citizen/employer/accountant) - Data entry (could be manual rekeying or could be full auto through online) - Calculation of tax due (could be HMRC, could be citizen using online) - Back end processing, adjustments to tax codes etc - The FT article mixes up PAYE (which is entirely an employer process) and Self Assessment (which is entirely an individual process), although both can and do involve accountants (or payroll bureaux etc). I vaguely remember there being something like 27 million people in the working population, of whom 2/3 are PAYE and 1/3 are Self Assessment. So if a quarter of tax codes are wrong, it could be in this proportion or, more likely, it could be mostly in one or the other (and I'd take a guess that PAYE is more likely wrong then Self Assessment - more later) - There are plenty of opportunities for error in the process. For instance: - The legislation may not cover all cases, or may put people in one category that should be in another - The data provided by the citizen may be wrong, accidentally or deliberately - The form filling process may lead to mistakes - The form submission process may break and returns could be lost - Data entry may result in keying errors - Calculation could suffer if data entry is incorrect - Back end handling could lead to further errors, perhaps where there are manual steps - Tax law is necessarily complicated for two reasons 1) It has to deal with a huge range of cases, from the secretary employed at a big firm, to the builder running his own company, to the Welsh mac-using vicar who runs a business printing bibles on the side, to the company director who earns a million, has houses that he rents out and has investment income all over the place. 2) I remember at a meeting with the Paymaster General some years ago, I was told that for every page of tax law there were 49 pages of anti-avoidance law. Accountants are clever folks and they're always trying to save their clients money. Of course, if tax law were simple, you'd be able to avoid (not evade) tax yourself and you wouldn't need an accountant to keep things straight (so that would be a few thousand people out of a job) - Because things are quite involved and lots of data from multiple sources is required, I doubt that everyone gets it right - so they either skip bits of data that are important (the occasional off shore bank account perhaps), or miss things out deliberately, or lose a bit of paper (dividend forms for instance, especially in a world where more and more stock trading is done online) or just get their knickers in a twist (before or after wetting them). So much of the errors in tax codes is probably a data problem - Keying in stuff of paper is not much fun for anyone and is highly prone to error - so until we get 100% electronic from source to payment, we're going to be left with some proportion of errors - and possibly quite a big proportion. Sadly, error rates in electronic items (especially PAYE returns from employers can be high too - at one point, magnetic tape records had 60% rejection rates because software used by the employers was not producing quality data). A lot of the detail on this can be found in Patrick Carter's report on PAYE. The HMRC response to this in the 2002 budget FAQ is famously brief "Q. What has happened to Patrick Carter's recommendations on simplification? A. The Government is considering these proposals and announcements will be made as decisions are taken." - The nice thing about government all the numbers are big - and they make for great headlines. There used to be a page on the IR website that showed how much tax was collected from each of PAYE, Corporation Tax, VAT and so on. I can't find it now, but the numbers are something like £100 billion in each case, give or take. So, if £2.8 billion of tax is uncollected, and that were all from PAYE that would be about 3%, if it were across all tax takes, it could be as little as 1% or maybe less (including, say, stamp duty). An error rate of between 1% and 3% might not be too bad and, if the errors break into several categories, then no single error is a big contributor. I'm pretty sure that credit card companies end up reserving that kind of amount for fraud with their cards (and that's one of the reasons why there's a big cost differential for processing credit card transactions versus debit card transactions) - There's an interesting wrinkle with tax. £2.8 billion of consumer tax going uncollected puts the same amount into the pockets of the population. What do they do with that? They go out and spend it and that generates VAT on the transaction, tax on corporate profits and so on. Trying too hard to collect every penny of tax that is due often leads to a greater involvement by more people in the black economy, and results in even more tax being lost. - On the other side of that, if £2.8 billion is going uncollected on one side, it's just as likely that £2.8 billion too much is being collected on the other. That would balance out the benefit from the bullet above and leave some people without money that they deserve. Am I defending HMRC here? No, not at all. There are vast numbers of broken processes in government - processes that have been broken for years because of bad thinking, poor design, terrible IT delivery, an over-reliance on paper, overly complex legislation, too much tinkering at the edges without structural reform and so on. Is the problem going to be fixed by trying to get right the 5% of returns that have errors in, or chasing after the 1%-3% of money that goes uncollected. Making the whole process simpler by making the data easy to get hold of, keeping it in standard electronic formats from beginning to end and speeding up the process so that people know what's going on will likely result in better changes. HMRC do need to take more accountability for mistakes - as does all of government. Finding out what went wrong and why is often a nightmare journey (witness my recent escapades with council tax and the valuation office which took me 12 months to resolve). Until then, Homer, you keep working on the XML, search engines and whatnot because that's what lets you pay your tax. Happy to see all the above stuff countered.
Posted by Alan at Thursday, February 09, 2006