DECC breaks National Statistics rules in order to hide its errors measuring PV

(I wrote to the UK Statistics Authority about the problems identified in this note. The Authority replied to me today (20.11.2015), saying it had asked DECC to use the correction mark where it had made changes in its statistics and also to make it possible for researchers to check how numbers had changed in running data sources such as the Renewable Energy Planning Database. DECC admitted its failure to properly identify its statistical changes in PV penetration because of an 'oversight'). 

The UK solar industry has been crushed and widespread bankruptcies have resulted. Part of the reason for the slaughter of PV seems to have been a perception in the Treasury that DECC had lost the ability to measure retrospectively the amount of photovoltaic capacity that has been installed, let alone control the future growth. For example, between June and September 2015, DECC increased its estimate of the UK’s installed PV capacity as of March 2015 by 1.2 GW, or over 15%. The yearly subsidy implication of this single error is almost £100m. So the Treasury was probably right to be concerned.

I looked today at DECC’s main publications that cover solar. There are at least four.[1] They show a consistent pattern of reacting to the problems measuring solar growth by blocking public data and refusing to tell researchers when substantial statistical revisions have been made. DECC has systematically broken many of the rules governing the release of government statistics over the last six months in order to disguise its failure to keep up with the growth of PV.

DECC has erred in four key respects

1)      Its publications have systematically underestimated solar growth over the last year (in retrospect, not just in prospect)

2)      DECC’s various publications today continue to publish very different figures about the amount of solar PV in the UK. Given that these estimates come from the same central statistical team, this is almost impossible to comprehend.

3)      The department has reacted to its failures by amending its previous estimates across its main databases. However it has broken Office for National Statistics rules by failing to note that it has made these retrospective revisions. Someone looking at current research will not be able to see that the numbers for previous months and years have been changed. 

4)      It has simply blocked access to previous published editions of two key data sources. Anyone seeking to look at previous editions of these databases is automatically redirected to the most recent report and cannot see the original data. Once again, this is a serious offence against National Statistics rules.

All these changes appear to have taken place within the last six months. At the beginning of 2015, DECC was aware that the reduction in subsidies for big solar farms in April 2015 would produce a rush to complete projects. It failed to realise just how much capacity would be installed. Despite spending significant sums on private sector suppliers of information, it had little idea of how much new capacity would actually be connected. Its database of planning permissions was incomplete. So it has acted to reduce the access of outsiders to its key data.

I’ll expand upon these points.

1, Systematic understatement of growth. In June 2014, Energy Trends estimated that the UK had 6.8 GW of installed solar (large fields and small roofs) as at the end of March 2015 when subsidies were severely cut back for new installations. In September 2015, this figure was raised to 8.0 GW. We can all understand how this mistake was made. DECC doesn’t give permission for farms to be built, nor are developers obliged tell it of completions so it is reliant on collecting data from 3rd parties. (By the way, I think the UK is the only large country in Europe not to have a comprehensive central database of solar installations). Ofgem - which handles the admission to the ROC subsidy scheme - is laggardly and uncommunicative about the applications it has received for subsidies. Although the dramatic March 2015 failure follows multiple upward revisions to solar PV estimates over the last few years for this reason, DECC’s error is excusable; it simply doesn’t have access to the stream of information necessary to monitor whether an individual farm is completed or not.

2, The variation in the simultaneous estimates of different government statistical reports is much less easy to understand. Energy Trends now says 8.3 GW for June 2015. Solar Photovoltaics Deployment, another Department publication, says 7.8 GW for the same date, a difference of 0.5 GW, or a difference in cost of up to £40m a year. The Renewable Energy Planning Database gives a figure of about 5 GW. This number is so wrong that I believe I must have misunderstood it.

3, I can readily understand the first two problems but I am very shocked by the latter two. Perhaps you will see these latter two statistical failures as largely technical but I think they seriously erode good government.

Point 3 is this: when the UK government publishes a statistic that it subsequently amends, it commits to publicising that revision by marking the new number with an (r) mark. This says that the figure is different from what was previously published. In the case of the solar estimates, DECC continued to tag its revisions properly until the numbers of Quarter 1 2014. These particular numbers were revised upwards for four successive quarters from September 2014 to June 2015 and each revision is marked. But none of the much more important revisions to quarterly estimates since the figures for Quarter 1 2014 have been tagged.  A user will not know the numbers have been changed by DECC.

For example, the shift upwards from 6.8 GW to 8.0 GW for March 2015 figures is untagged. And neither are any of the revisions to PV data over the past year. Without serious research, no-one can know how serious the errors in retrospective estimates have been.

This is what DECC says about correcting statistics.

Data that has been revised will be indicated by an ‘r’; the ‘r’ marker will be used whenever data has been revised from that published, either in printed form or on the Internet.

When it comes to PV, this has simply stopped happening, across all the many databases. I believe DECC has systematically broken the rules set by the independent Office for National Statistics.

4, Perhaps we can even comprehend point 3. No civil service statistician will want to flag errors of the size and expense of March 2015. It will have been easy to forget to make it clear to readers that the numbers have been changed. There is no such excuse for point 4, the removal of previous data series from the internet.

Until recently, if I visited the DECC site I could see the monthly evolution of the series it publishes called Solar Deployment Trends. I could compare the estimates of July 2015 and August 2015 for the amounts of PV in, for example, April 2015.

No more. Although these databases are still written down on the long list of DECC’s published statistics, the Internet links have been killed. Anyone clicking on the item is now redirected to the most recent estimates. The file has gone from the web. The government is now making it impossible for researchers and journalists to see what it said were the figures for installed PV just a month ago. Simply put, this is a reversal of all recent government policies on open data.

I know why DECC has done this. The figures in this particular database were often highly – and obviously – inaccurate. (For example, I wrote to DECC in April asking for an explanation of a 1 GW apparent inconsistency. Almost a month later I got an explanation which avoided the issue).

Similarly, the government no longer allows access to past editions of the Renewable Energy Planning Database, possibly the most flawed government data series I have ever seen, despite being assembled at considerable expense by an outside company.

As I wrote earlier, perhaps my concerns are inconsequential. What does it matter if the government is cagey about its past errors of estimation and removes the most obviously flawed reports?

I think the effect on energy policy, and on the process of government more widely, is potentially highly detrimental. The last year or so has shown that solar PV can be installed in huge volume in the UK at a speed that is almost uncontrollably fast. Massive forecasting errors were made – by almost everybody.[2] (My own flawed estimate for was that at end March 2015 UK solar capacity was less than 7.3 GW, a mistake of about 1.0 GW).

During a trade show in Beijing earlier this month, China is reported to have an announced a doubling of its targets for PV in 2020. The country now aims for 150 GW of solar within five years, up from a previous target of 75 GW. The lesson of last year is that the UK could easily have achieved a similar ambition, pro-rata to population. Instead, we try to hide the evidence of how unexpectedly successful we were, largely because the measurement errors are embarrassing and destroyed DECC’s credibility in the eyes of the rest of government. Much of the rest of the world has pretty much decided that solar PV will be the basis of national energy systems just as the UK throws tens of thousands of skilled solar installers into unemployment.







[1] Digest of UK Energy Statistics, Energy Trends, Solar Photovoltaics Deployment and Renewable Energy Planning Database.

[2] Some solar experts, such as Ray Noble, were much more accurate.