Bumping Up the Figures

.
I remarked the other day on the composition of the PAS database with regard to the percentages of different types of finds. This is not the only odd feature. Readers might find it entertaining to use the statistical tables in the revamped PAS database to plot out its growth as a histogram (you'll need a big piece of paper if you do it by hand). Plot both "records" and "finds". Quite startling. The curve starts off gently expanding, then after march 2003 it climbs steeply. Then after March 2007 it goes wild. "Brilliant success" the histogram screams. "What on earth...?" says a veteran PAS watcher.

The notion of the "objects" in the database is self-explanatory. That is the statistic that is most often quoted in the PAS propaganda of success. But its not the most significant value. To illustrate this, let's take a kiln waster, a clearly overfired and distorted Roman potsherd. On its own it is not necessarily significant. Find ten of them in an area of a field with a couple of lumps of hard fired clay scattered in an area five metres diameter, they become more so, collect forty of them from the same area and there is little doubt that they should be interpreted to mean there had been a Roman kiln at that spot, and possibly remains of it survive in the subsoil. If a fieldwalker collects four hundred of them, they'll need a bigger box to put them in, but in terms of the basic information, the location of a Roman kiln producing a certain type of pottery, collecting four hundred sherds could be data overkill. In terms of a record, two give all the information that is needed to put that find on a map:
"forty Roman waster sherds of Alice Holt fabric three vessel type Jeffries 43a and 56b found at NGR XR 12345 67890 by fieldwalker Ivor Lookaround in December 2011 in a discrete area five by five metres, deposited in Hangthemall Museum acc no. 66789.11" and "sample of hard fired clay from dense scatter in a discrete area five by five metres found with Roman waster sherds at NGR XR 12345 67890 by fieldwalker Ivor Lookaround in December 2011, deposited in Hangthemall Museum acc no. 66790.11"
As putting an annotated dot on a map is concerned, counting the number of objects taken out of the field and put in a finds bag is only secondary to that.

Thus in terms of documenting archaeological context, the statistic in the PAS database that should be noted is the number of records, not the overall number of objects they comprise, one pair of records of finds indicating a kiln discovery is the same pair of records whether forty objects were taken out of the field or four hundred.

So why does the PAS tend to quote the latter? Because it is the bigger number. Today there are, it says, 441,646 records, so 441646 reports of archaeologically useful information recorded as a result of PAS outreach but "698,639 objects". So that's nearly seven hundred thousand "objects" - objects what? Seven hundred thousand objects taken out of the archaeological record, but where are they? What actually is the value of that piece of information? Actually none whatsoever. Its a big number that is intended to sound impressive and "good value for money".

But is it good conservation? Well, for some years the Heritage Action Erosion counter has been ticking away with (what I am convinced is) a quite conservative estimate of what the archaeological record of England and Wales is losing month by month due to the activities of artefact hunters with metal detectors. Since the PAS was set up, it predicts that 4,189,936 recordable items have been taken out of the soil. In that time, the PAS has made 442 000 records (but nearly 700 000 objects). So that's a shortfall of 3,489,000 objects since the PAS began. Three and a half million objects gone without record from under the nose of the PAS and its legion of uncritically enthusiastic supporters at home and abroad.

It could be worse of course, I mean, look at the figures from February 2010: Total records - 289685, total objects recorded - 456806. So how has this suddenly jumped up in the past year or so to current levels? Are metal detectorists at last flooding to PAS offices with car boots-full of individually bagged artefacts each with a ten figure NGR? Are we at last seeing the fruits of all that expensive dedicated liaison and head-patting, back-slapping partnership and camaraderie with artefact hunters?

Sadly, it seems there are grounds for believing that nothing could be further from the truth. There are instead two other processes in operation here which only superficially create the impression that collaboration with artefact hunters has improved since the early days. Indeed, one might ask the PAS why they do not present the data in a manner more transparent in order that such comparisons can be made between - say 2003-2006 and 2008-2011. Now I could suggest "one reason", but it would be nice to hear their side first, wouldn't it?

The first set of processes that can be demonstrated involves the way the "number of objects" is being bumped up. On the graph for the first decade of PAS operation the curves for the "number of records" and "number of objects" climb up more or less at the same rate, gently diverging from each other as a record consists now and then of a dozen or so potsherds or flints in a bag. Then about March 2007 they start diverging from each other by an increasingly large degree. What is going on? Are the bags of potsherds getting bigger? Are artefact hunters being asked to pick up as many potsherds or flints (because, it's not iron nails and iron slag is it)? I must admit I puzzled over this one for some days when I noticed it, and of course one can search in vain the PAS website for any kind of explicit explanation of what the figures mean beyond "wottalotta stuff we've seen".

But there is one inadvertent clue. Not even on the PAS website, but a chance - offguard?- remark on Facebook no less. It's the usual wottalotta-type remark:
Since we relaunched the database/website last March 4335 people have found and reported 132,274 objects (skewed by one hoard of 52,503 coins!). Database users view 16.12 pages per visit and stay for 13mins 30 seconds. Thanks to you all for making it successful. Now bring your friends and show them what's out there.
What is a 52000 coin hoard doing on the PAS database? The PAS database is for non-Treasure finds, isn't it? Well, the clue is that in March 2007 within the structure of the BM the Portable Antiquities Scheme was formally linked to the Treasure Unit, which had formerly functioned semi-independently (Roger Bland being head of both). What seems therefore to have happened is that some, or maybe all, Treasure statistics are now being added to the PAS database as PAS-data. The Staffordshire hoard is there for example. That is why after March 2007 the two curves diverge so strongly, the "number of objects" statistics are now being supplemented by coin hoards containing several thousand items apiece. In no way can the figures presented as "PAS success" on the PAS website be attributed alone to the success of PAS outreach (Treasure finds have to be reported by law, their reporting is a totally different phenomenon from the voluntary reporting of non-Treasure finds by responsible metal detectorists). What the ACTUAL detailed figures for the latter are is anyone's guess, the PAS has for the past couple of years seemingly gone out of its way not to present them to public scrutiny.

But that does not provide the whole explanation for the amazing jump of the past year or so in PAS database statistics. To suss that out, compare the database figures for Friday 19th March 2010 and those for Monday 22nd March 2010, up from 291559 records of 459630 objects to 382 303 records of 550 374 objects overnight. Coin elves in the BM? A very big commercial metal detecting rally? Actually this was the entry onto the PAS database of Oxford University's Celtic Coin Index, data gathered independently of the PAS since the 1960s. I am surprised Canadian celtophile did not notice the inordinately high percentage of "Celtic coin" finds on the PAS database search I discussed here earlier. This is the explanation why now it is much higher than the **% pre-CCI-data-insertion. It seems that about the same time another externally-compiled database was added to the PAS database, the Iron Age and Roman Coins of Wales database compiled at Cardiff University. The addition of these two databases is responsible for the relatively large jump in numbers of records made in 2010 and (since both external databases also contain records of hoards) the "number of objects" curve to give the current inflated total.

While to some degree some may argue that these databases - being records of "portable antiquities" - belong in the PAS database (I don't), it is without question that the lack of differentiation of these different types of statistics in the one dataset renders it completely useless as a means of assessing to what degree the PAS is achieving success in their outreach to metal detectorists and to what degree it is being successful (or not) in mitigating the information lost through unrecorded artefact hunting. This is not an insignificant question, for it was to achieve this that the PAS was set up, and to which end thirteen million quid has now been spent on it in direct funding alone.

It would be honest and transparent of the PAS to provide some sort of public explanation of the contribution individual components of their database are making to its overall shape, what it represents and what it does not represent, allowing the fuller analysis of these figures concerning the effects of expenditure of public funds. Otherwise it is simply creating a false impression of the degree to which artefact hunters are voluntarily reporting their finds to the Scheme.

We might address a direct question to the PAS what these statistics actually are, and why they choose not to make them visible in the public domain, but I really think we will get nothing in the way of an answer from them. My guess is that they are well aware that the figures are not really very much in favour of the notion that they are achieving anything like a satisfactory result in the way of mitigation of the erosive effects of ten thousand of their "partner" metal detectorists stripping out selected geegaw goodies from the archaeological record for collection for personal entertainment and profit.