I’m using the next few days to unpack some of those ideas you have milling around in the back of your head that might either come under the category of Good Idea or, alternatively, What Were You Thinking…

Let’s look today, then, at how HMRC might organise the work it does to counter tax avoidance and evasion in direct taxes – the kind of stuff that I used to do as an Inspector, back in the 20th Century.  In those Olden Days of Yore people and companies used to send in their accounts (on pieces of paper!  I know!!) and then someone used to look at them and decide which category they came into.

There were three categories, I seem to recall.

• A for Accept.  Someone had a quick look, couldn’t see anything that looked like a problem, so sent the file off for processing.  Most accounts came into that category (because – and let’s not lose sight of this fact – most people are actually honest).
• R for Review.  That meant that someone like me looked at some of the figures in a bit more depth.    Bear in mind that I was never actually a very GOOD inspector, technically, but even someone like me was expected to bring in around a million a year in adjustments from this kind of bread-and-butter challenge.
• E for Examine.  Not just one or two figures but the whole basis of the accounts were considered for some reason to be questionable, so the whole thing was investigated or “examined” in depth.

Yes well, that was a long time ago, and I imagine we’re much better off now, looking systematically at risk across the piece rather than relying on the prejudices, talents and know-how of individual inspectors acting alone.  What I’d like to propose, though, is the next step, a new three category system.  (Remember, now we have self-assessment, so the old “A” category looks after itself.)

So my first category is I for Inspection.  I propose that a reasonable proportion of the accounts submitted to HMRC should be audited, with the cases chosen for inspection on a random basis.

Because personally I’m not convinced – admittedly on the basis of no evidence except a gut feeling – that risk-based selection gives a better result than random selection.

HMRC is, understandably, cagey about any figures it might have on the comparative value of random investigations but there is a clue in their tax gap paper, where they say

Most compliance work is risk based, and it can be difficult to use the information gained from such enquiries to assess revenue losses from other taxpayers. However HMRC also undertakes random enquiries, the results of which can be used to extrapolate figures for the rest of the population.

So they still do some “random” enquiries, but of course they’re not going to tell us how many or what the results are, because they have enough problems with people trying to “game” their systems.

But think about the horsemeat scandal.  Was a risk-based inspection system really the best way of guaranteeing the food chain?  Or would we have been better served by someone brandishing a clipboard marching into food processing and distribution centres on a random basis and saying “your number’s up.  Give us a sample of everything…” and then having a proper look?

My first proposal, then, is to have a big chunk of compliance activity take place on a random basis.  So there would be no heat about it, no suggestion of wrongdoing – your number would just be up.  It’s your turn to be audited, mate, sorry – I’ll come round on Thursday to look at all your books and records, all right?

Next: E is for Evaluate.  Someone who wants to come clean before they’re either inspected or investigated can of course come forward and come clean.  You’d call those “evaluation” cases, where you would have a team to take their confession, check it was complete and accurate, and reach a settlement including appropriate financial penalties.

Why is this post called “random pie” though?  Random is pretty obvious – I want to rebalance compliance effort from “risk” to “random”, and PIE is of course an acronym.  We have I for Inspection and E for Evaluate… how about P for Prosecute?  Because one of the things I think was missing from the old system was prosecution.  Yes, the old Revenue used to prosecute a handful of cases, but the discrepancy between the amount you can get prosecuted for if you’re a benefit fraudster  and the level of tax fraud that is settled without any prosecution is striking.

So here’s my suggestion.  Pick a number – almost any number – and stick with it.  Call it the “prosecutable offence limit”.  If you have fiddled your tax – or fiddled your benefits – over the limit, you won’t get a civil settlement but a criminal one.

So there would still be “intelligence led” and “risk based” enquiry work, but it would be conducted differently.  Intelligence led investigation no longer leads to civil settlement but to investigation under PACE and to prosecution.  Yes, of thousand pound tax fiddles, or at least of fiddles over a common threshold to be agreed – if someone’s prosecuted for a thousand pound benefit fiddle then equally someone who’s committed a thousand pound tax fiddle – at least one that’s evidenced to the criminal standard – should be prosecuted.  Why not?

Prosecute where you have evidence, Inspect a random number to keep the system honest, and Evaluate people who confess voluntarily.  Random PIE.  Why not?

This post is the second of ten posts I intend to write between now and Red Nose day.  If you feel like supporting me with a morsel of sponsorship, my JustGiving page can be found here.  And thank you!