1. We present a Northern-Ireland-electricity-cost-puzzle. In our view, the household cost reported by the Office for National Statistics (ONS) is unlikely to be as accurate as otheres timates from the Department for Business Energy and Industrial Strategy (BEIS) or Utility Regulator (UR). We are unable, however, to confirm the source of bias affecting ONS’ data.
2. On the one hand, according to ONS, electricity costs in Northern Ireland (NI) are higher than any other UK region.
Figure 1: Weekly spend on electricity per household across UK Regions, 3-year period toMarch 2018
3. On the other hand, according to BEIS, electricity costs in NI are, in fact, lower than anyother UK region.
Figure 2: Electricity charges across UK regions, 2018 calendar year, average supplier,including VAT
4. The Utility Regulator’s (UR’s) Quarterly Transparency Reports (QTRs)1use the same method & process as BEIS and hence report a similar conclusion – that electricity costs inNI are in fact lower than other regions in GB.
5. In this paper, we explore reasons for this apparent contradiction.
6. The ONS methodology is based on household surveys, where households are asked howmuch they pay for electricity, using (if possible) last bill/payment. The relevant‘bill/payment’ period is also captured, which ONS then use to calculate a ‘weekly cost’.
7. In contrast, the BEIS methodology is based on information received directly from supplycompanies, via a quarterly survey. BEIS explain the methodology as follows:
“The suppliers provide figures for each tariff (unit costs, standing charges,split levels, discounts, dates of tariff changes and number of customers),splitting the tariff information by payment type and region. Data is receivedas part of a quarterly template, sent out to energy suppliers shortly after theend of each quarter. All information received from suppliers is qualityassured by BEIS prior to publication.”
8. Not only do the methodologies differ, the output also differs. The ONS methodologycaptures what households believe the household cost is, including the quantity of energyconsumed. In contrast, the BEIS methodology reports what suppliers say the charge per unitis, excluding the impact of consumption. Therefore, the reported values, from ONS and fromBEIS, may be consistent, if consumption-per-week explains the difference, as demonstrated in the following equation:
Weekly cost (WC) = Unit Cost (UC) * consumption-per-week (CPW)
9. Using the UC reported by BEIS and the WC reported by ONS, the balancing figure for CPWcan be derived, on the assumption that both UC & WC are consistent, by dividing WC byUC.
Table 2: Implied Consumption minus Reported Consumption. A puzzling difference.
10. The implied domestic consumption can now be compared to reported domestic consumption, using data published by UR2, as follows.
11. Clearly, there is a material inconsistency in, or misinterpretation of, the available data on electricity consumption and/or costs. The explanation(s) for this puzzle must be:
a. We somehow misinterpret the data,
b. BEIS understate electricity unit costs,
c. UR understate electricity consumption, and / or,
d. ONS overstate weekly electricity costs.
12. The average difference in Table 2 column D is 27%. But because we are not sure whether the difference is based on consumption errors or cost errors (or both!), we can call this puzzle, ‘the missing 1,000kwh per year’, or ‘the missing £150 per year’
13. Combining different sources of information is risky, and we cannot rule out the possibility that simplifying assumptions could explain this puzzle, at least in part. To test this, we address several assumptions underpinning Table 1 and Table 2.
14. Thus far, our presentation implies that the periods are directly comparable - however this is a simplification. Firstly, the ONS data is based on 36-month periods, ending in March each year. So, in Table 1 column A, the value for 2018 (£12.60) represents the average weekly cost for the 36-month period ending March 2018. Secondly, the BEIS data relates to 12- month periods ending 31st December each year. So, in Table 1 column B, the value we use for 2018 (15.63pence per kwh) relates to the 12-month period ending December 2018
15. However, comparing 3-year periods with 1-year periods, is unlikely to explain the puzzle in a material way. If it were a material factor, we would expect to see differences that were sometimes-over-sometimes-under, during a sufficiently long time-series of data3 Similarly, if mismatched periods (ending December rather than ending March) were driving the differences, we would expect temperature-corrected-consumption to be material – however using BEIS (UK level) data for over the relevant periods, temperature correction adjustments are usually smaller than 3%.
16. Separately, a further assumption in Table 2 is that consumption values, are directly comparable. This is not a perfect assumption because the UR data is based on connections whereas the ONS data is based on households. If there are more connections than households, then the consumption per connection (Table 2 column B) is understandably lower than the consumption per household (Table 2 column A).
17. However, in our view, we do not think this is a material issue. UR’s connection numbers are very similar to the number of households reported by ONS. In effect, the connection-tohousehold ratio is small. For example, in 2017 the ratio is 1.01 (i.e. a 1% difference). Although the ratio is larger for previous years. Hence, overall, we don’t think this distinction is material enough to explain the puzzle
Table 3: Connection-to-household ratios are small and hence don’t explain the puzzle
18. Table 1 column B refers to ‘unit costs’ but if these exclude fixed charges, it could contribute towards the puzzle. However, the BEIS methodology confirms that “unit costs reflect the prices of all suppliers and include standing charges”.
19. Further, Table 1 column B does not reveal whether these ‘unit costs’ relate to credit customers, direct debit customers, or prepayment customers (or some combination thereof). If the BEIS data is based on, say, direct-debit customers (who pay less) instead of creditcustomers (who pay more) this may help explain the puzzle. However, again, the BEIS methodology is reasonably clear:
“The average bill is equivalent to the total revenue divided by the total number of customers. For each tariff, the total number of customers in a year is equivalent to the average number of customers across the four quarters. For each tariff, total revenue is equivalent to the average number of customers multiplied by the sum of the bills in each of the four quarters.”
20. To support this, BEIS publish the average unit cost for each payment type (which BEIS refer to, loosely, as ‘tariffs’) and each region. The charge for credit customers is higher than direct debit customers (by 7% in 2018, and less in preceding years). However, the BEIS publication shows the overall average unit cost across all payment types, which is what we use in Table 1 column B to represent an ‘average customer’. In any case, even if we used the charge for credit customers, (more expensive than direct debit and prepaymentcustomers) it would not fully explain the puzzle.
Figure 3: Payment methods do not seem to explain the puzzle
21. One part of the BEIS methodology is worthy of further investigation, in terms of what ismeant by “No allowances are made for introductory offers or non-cash benefits that may beavailable from suppliers.” However, if we assume that introductory offers and non-cashbenefits would reduce the BEIS unit costs, this would increase the puzzle, rather than explain it.
22. Lastly, BEIS confirm that its data is inclusive of VAT, ruling out a potential 5%understatement.
23. UR’s Quarterly Transparency Reports are, in terms of both connection numbers and consumption per quarter, based on information from Northern Ireland Electricity Networks (NIEN). NIEN’s consumption data is based on estimates, because not all electricity meters will be read simultaneously on 31st December each year. Therefore, it is possible that NIEN underestimate true consumption, particularly for the quarter ending 31st December, when consumption can be higher than the previous three quarters (assuming NIEN’s estimates are impacted by smoothing).
24. On this basis, underestimation is more likely for quarters ending 31st December, but by the same logic, it is also more likely that other quarters are overestimated. On an annual basis, the net effect should result in no systematic bias.
25. It is therefore unlikely that underestimation would occur every year for four consecutive years. To explain the puzzle in Table 2, NIEN’s estimates would need to systematically and materially underestimate true consumption. Underestimation is more likely to occur when sales of electricity increase, as has been the case in NI (unlike mainland GB, where electricity sales have fallen by 11% (England and Wales) to 22% (Scotland).
Figure 4: An index of electricity sales (volumes using TWh) to consumers, 2002 = 100
26. Nonetheless, in the absence of anything more tangible, systematic underestimation by NIEN(and hence within UR’s QTRs) seems, in our view, unlikely.
27. We now address the key components of the ONS approach.
28. ONS explain its sampling methodology4 as follows:
“The LCF (Living costs and food survey) sample for Great Britain is a multistage stratified random sample. Addresses on the Postcode Address File with ‘small user’ postcodes are used as the sample frame. Postal sectors are used as the Primary Sampling Units (PSUs), with 18 addresses selected from each PSU to form the monthly interviewer quota. A total of 638 PSUs are selected annually after being arranged in strata defined by Government Office Regions and two 2001 Census variables: socio-economic group of the head of household and ownership of cars. In Northern Ireland, the companion survey to the Great Britain LCF is conducted by the Central Survey Unit of the Northern Ireland Statistics and Research Agency (NISRA). A systematic random sample of private addresses is drawn from the Land and Property Services Agency’s database.”
29. A random sample of private addresses by NISRA should avoid sample bias affecting its reported numbers, although it is not clear from this quote whether the random sample is based on postcode areas. Consider now that Northern Ireland may have a higher proportion of rural, and possibly larger, households, there is a potential that a fair sample produces a different type of household, perhaps with larger houses and/or with more occupants, than other GB regions. However, even if this is the case, it would not explain the puzzle, because we have compared average NI consumption using different sources. If the sample is truly random, the ONS value should be close to the average, but clearly the average implied consumption and the average reported consumption are very different.
30. Prior to 2016/17 only 150 households from NI were captured in the ONS survey. Since 2016/17 however, NISRA explain that the sample size is “boosted to improve precision” – it now captures approximately 400 households per year. However, NISRA also explain that 150 households would suffice and would be comparable with sample size proportions inother GB regions (0.02%). The boost in the sample size was requested (and funded) byanother department5and may, in due course, be reduced back to 150 households. Note alsothat ONS report 3-year averages for each region rather than annual values. The motivationfor this, NISRA explain, is to further improve the accuracy of the data. Further, NISRA alsoexplain that the approach in NI is likely to be superior to GB, because the systematic randomsample is not based on postcodes, as is the case in GB, and is therefore less restricted whenobtaining a reflective sample of households across population areas and house sizes. In thisrespect NI data is, in NISRA’s view, equally (if not more) reliable than other UK regions.
31. However, this does not necessarily mean that the sample is equally as accurate as other sources of NI data, which are more focused and contemporaneous.
32. Nonetheless, in these respects, it is difficult to see how sampling size/bias could explain the puzzle
33. Statistical estimates, such as those produced by ONS, are inherently uncertain. One measure of uncertainty in this regard is the ‘standard error’. Unfortunately, it appears that, for each GB region, ONS do not publish standard errors for individual expenditure categories.
34. At the overall GB level, the standard error for electricity costs is small (1.1% for year-end March 2018). In effect, this means that ONS is 95% confident that average expenditure on electricity costs, at the overall GB level, is within a small boundary of approximately £0.50 per week. Intuitively, standard errors for a homogenous product like electricity, borne by all households, are smaller than other household cost categories.
35. However, we would expect the standard error for region specific estimations, including electricity costs for NI, to be much larger, given that NI is a subsample of the larger dataset which will likely lead to a larger variability in survey data.
36. Unfortunately, NISRA cannot share the source data with us from its survey responses. With this information, it should be possible to confirm whether statistical uncertainty could explain the puzzle. NISRA explain that standard error calculations are not simple to calculate for individual categories per region because the raw data is weighted and adjusted, prior to publication by ONS. However, NISRA agreed to request this data from ONS, to help us explore whether this could explain the puzzle
37. In the meantime, we estimate that the standard error would need to be at least 7%, for us to conclude that the ONS values are not, in statistical terms, ‘significantly different from the puzzle implied weekly cost’. In other words, if the standard error is >7% for NI electricity costs, then the true weekly cost could be around £2 lower per week, hence explaining the puzzle.
38. ONS explain its fieldwork approach6 as follows:
“The fieldwork is conducted by ONS in Great Britain and by NISRA for the Department of Finance and Personnel in Northern Ireland using largely identical questionnaires. Differences between the two questionnaires reflect the country-specific harmonised standards for ethnicity, nationality and national identity, and the different systems of local taxation used in Great Britain and Northern Ireland. Households at the selected addresses are visited and asked to co-operate in the survey. In order to maximise response, interviewers make at least four separate calls, and sometimes many more, at different times of the day to households that are difficult to contact.Interviews are conducted by Computer Assisted Personal Interviewing (CAPI) using laptop computers. Respondents complete a face-to-face interview and each individual aged 16 or over in the visited household isasked to keep a diary of daily expenditure for two weeks.”
39. Given that the ONS methodology is dated 2012, we sought clarification from NISRA with regards to whether it remains an accurate reflection of what has taken place since that date. NISRA confirmed that interviewers attended households, conducted CAPI using laptop computers and collected 2-week diaries (although electricity costs are not captured in these diaries). Further, to incentivise participation, NISRA, like ONS in GB, provide £20 remuneration for adults and £5 for children (under 16-years-old). NISRA further explain that incentive-rate-experiments are taking place in GB, using £40 rather than £20, to test response rates. Currently, however, in NISRA’s view, there is no systematic fieldwork or incentive difference between GB and NI.
40. To test whether there may be a time-of-the-year impact, resulting from high costs in December / January for example, we asked NISRA to confirm the calendar month that interviews normally take place. NISRA confirmed that interviews and surveys are conducted throughout the year, and hence unlikely to be affected by time-of-the-year effects. Further, we asked NISRA if respondents may have ‘rounded-up’ their responses, say from ‘£35 per month’ to ‘£40 per month’ for convenience / ‘nearest £20’ purposes. However, NISRA confirmed that interviewers are trained to avoid such bias, by seeking ‘bills/statements’ proofs and ‘pounds-and-pence’ accuracy from respondents. We also asked NISRA if, given the small incentive of £20, there may be insufficient time available for respondents to find/source their electricity bills/statements, particularly given that the survey attempts to capture almost 200 other cost categories, such as food, clothing and recreation. NISRA agreed that the surveys are onerous exercises, taking up to 90 minutes with participants.
41. Unfortunately, NISRA could not share a copy of the computer-based survey with us, so we are unable to develop a deeper understanding of its methodology.
42. Nevertheless, based on our review of the published methodology and our discussions with NISRA, the fieldwork in NI is not, in our view, demonstrably biased in any clear way that would overstate true weekly costs. Can we use Budget Energy data as a cross-check?
43. In the absence of solving the puzzle, we can nonetheless calculate an independent weekly electricity cost that will indicate the true average weekly NI electricity cost. Budget Energy Limited, an electricity supplier in NI, provide us with a suitable cross-check because it has a large domestic customer base (more than 50,000 domestic customers, representing approximately 99% of its total customer base on average, over the four-year period ending June 2017). Further, a large proportion of the electricity it sells is to domestic customers (97% over the four-year period ending June 2017). Put another way, only 1% of its customer base, or 3% of the electricity it sold, relates to non-domestic (industrial / commercial) clients.
44. From Budget Energy’s financial accounts, we collect customer numbers and total revenues. Dividing total revenue by customer numbers produces an annual average revenue/spend per customer (and dividing again by 52 weeks produces a weekly average revenue/spend per customer, for more than 50,000 domestic customers)
45. Again, compared to the ONS information, the Budget Energy cross-check implies a much lower electricity cost. The puzzle is a similar magnitude, of more than 25%, to that presented in Table 2, thus reinforcing our characterisation of the puzzle as ‘the missing £150 per year’. Note also that the Budget Energy revenue/ spend per customer is near (or less than) £600 per annum, which is consistent with the BEIS data presented in Figure 2.
46. Note, however, several limitations with this cross-check. First, the periods do not exactly align, as Budget Energy’s financial accounts relate to 12-month periods ending June each year, whereas the ONS values relate to 36-month periods ending March each year. Further, Budget Energy, as a new supplier, needed to offer a cheaper tariff than other NI suppliers to attract new customers. In addition, these energy-savvy-switchers may consume lesselectricity than the NI average.
.jpg)
MCC’s submission on the AEMC’s electricity pricing review: nine strategic trade-offs, protecting vulnerable consumers, pairing prices with technology, and clear obligations on retailers.
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
Block quote
Ordered list
Unordered list
Bold text
Emphasis
Superscript
Subscript
I was delighted that MCC's work was completed on time, and within budget, helping us deliver important changes and improvements, to the benefit of our stakeholders. MCC's report is published on the CCC website.
I am delighted to recommend MCC Economics. Specifically, I worked closely with PJ, who helped us with our Nuclear and CCUS projects. PJ helped us develop new policies and answer questions from our stakeholders. His support helped us deliver important changes and improvements, to the benefit of our stakeholders.
MCC Economics has helped us better understand the most important issues for our stakeholders, including: charges, shareholder returns, debt payments and inflation impacts.
I am delighted to recommend PJ and his team at MCC Economics. We've been working together on National Policy Statements to help meet net zero targets for 2030 and 2050. We initially appointed MCC Economics to support us on offshore wind consultation analysis and have recently reappointed MCC Economics to undertake a larger consultation analysis role across all sectors, including hydrogen, CCUS and networks. I can confirm that PJ and his team have shown excellent spreadsheet skills, alongside very good project management, planning and analysis skills, helping us deliver important changes, and continuous improvements, to the benefit of our stakeholders.
I am delighted to recommend PJ and his team from MCC Economics. They helped us with our price controls for Heathrow airport and for NATS (En Route) plc (the air traffic services provider). Specifically, the MCC team helped us deliver important changes and improvements to our financial models and supporting policy documents, to the benefit of our stakeholders.
I am delighted to confirm that I worked with PJ on a retail project in 2015. The project helped stakeholders understand electricity costs and charges. Specifically, the project helped us explain to stakeholders, internally and externally, why electricity charges differed across the regions (GB, NI & Ireland). PJ was a key member on the project team, which helped deliver changes and improvements in the understanding of energy retail.
.png)
Evidence-based analysis for transparent, defensible, and effective decision-making.
© All rights reserved – MCC Economics 2026.
