Research study quantifies energy theft losses


Research study quantifies energy theft losses

Pundits claim two Scots fighting over a penny found on an Edinburgh street led to the development of copper wire. The malleable quality and conductive capability of copper led Thomas Edison to use the metal in a number of his inventions around the turn of the 20th century. One of Edison’s major contributions to society was the generation and distribution of electricity. It is reasonable to assume that the theft of electric service began shortly thereafter.

More than 100 years have passed since Edison introduced the rudimentary elements of electrical service, yet to date practically no one in the electric utility industry can estimate the amount of financial loss to theft and diversion with any degree of certainty. Put another way, no systematic examination has been conducted that would allow the precise articulation of what percentage of revenue is lost to thieves.

“All the data estimating theft are anecdotal, and many utilities don’t have any sense of the magnitude of their losses,” said Ralph Abbott, president of Plexus Research Inc. Plexus Research is gathering data that is expected to quantify losses at commercial services.

The May 2000 issue of the International Utilities Revenue Protection Association (IURPA) News carried a response to a letter to the editor from George W. Davis, president of Inner-Tite Corporation. The letter was a request for facts and figures regarding EEI (Edison Electric Institute) survey estimates and other similar types of information. Davis pointed out that several EEI reports, memoranda and surveys had been completed over the years, in particular in 1977, 1985 and 1990. Other articles and surveys were mentioned, including a 1999 EPRI article and the 1997 Canadian Electricity Association survey. Available information regarding energy theft continued to be subjective, at best.


United Energy of Australia provided the first real breakthrough in data collection, interpretation and extrapolation. Its revenue protection unit conducted a study designed to detect faulty meters. A total of 6,210 accounts were interrogated, and 427 meters selected for field-testing. Of those, 134 meters were found to have some sort of defect. United Energy was thus able to determine that 2.16% of its meters were faulty, which represented a potential loss of $4.5 million (Australian) to the company.

This research was important beyond the determination of loss. It demonstrated a real effort to quantify loss with something other than an extensive survey conducted by an industry-related organisation.

The Canadian Electricity Association’s Extent of Energy Diversion on Customer Premises for Canadian Utilities was another important research study, in that it provided a benchmark by which to judge everything that followed. Its stated purpose was “to statistically assess the incidence of energy diversion at several utilities across Canada.” The overall approach to the research of energy theft, including the sampling strategy, was both academic and practical. The study’s conclusion, however, did not match the sterling achievement of the project itself. “Deviations which will certainly lead to diversion are definitely occurring across Canada. The average rate for these deviations is 1.36%. At this rate, on total sales to residential customers in Canada in 1993 of 132,327,911 MWh at 7.25/kWh, this accounts for a revenue loss of $130,475,320.” 

This is the point where the Canadian survey (and seat-of-the-pants estimates) falls short. For the 1.36% tamper rate (deviation) to equate to $130 million, theft would have had to occur for an entire year. Experience has shown this is not the case. More typically (based on actual billings) theft will occur for three or four months before a meter reader or field services technician spots an irregularity. Alternatively, on an account that is diverting power around the meter, a meter reader will note a zero consumption report into his Itron data-cap, and after 60 to 90 days a field inspection will be called for, or an SONP (shut-off for non-payment) audit will occur. Theft and diversions that last a year or longer do take place, but in our experience they’re the exception.


The revenue protection team at Arizona Public Service Company determined that combining the best research attributes of United Energy and the Canadian Electricity Association would provide the information we required, but that an additional step was needed. Both had provided excellent direction and information regarding the amount of various meter problems found in the field, and could cite specific percentages (United Energy 2.16% faulty meters, Canada 1.36% tamper rate). We felt we needed to take the next step and conduct investigations to determine the actual dollar loss associated with each meter found to have been tampered with, rather than settle for overlaying a tamper rate percentage on to total revenues. 

The goal of the research study at APS was, therefore, to determine the dollar amount of loss to theft and diversion. The first issue was to take a statistical sample of the entire meter population in our service territory (868,000 meters) to determine the extent of meter tampering. The APS Audit Services Department, specifically Cynthia Reed, CPA, was consulted for statistical direction. In order to determine the sample size the following assumptions were established:

  • Confidence Level – (the probability that the value obtained by a sample will not differ from the true value of the universe by more than a stated amount (precision).) 
    We selected 95%, giving us 95 chances in 100 that the sample was representative of the entire population.
  • Expected Error Rate – (the percentage of error (theft and/or diversion) expected to be found as the result of the research study.) 
    Determining an error rate can be difficult, as knowledge can be inadequate and incomplete, but reviewing the results of previous audits, and conducting preliminary surveys or a small pilot test, will help the process. The rate determined must be realistic; when in doubt using the higher end of the range or estimate is recommended. We estimated our expected error based on limited information and existing industry quotes. Using the high end of the spectrum, the Edison Electric Institute/Justice Department’s estimates, we selected an expected error rate of 3%.
  • Precision – (a range within which the true answer concerning the population characteristic under study should fall, at the specified confidence level.) 
    We selected a precision range of ±1.5%, thereby giving us the range of 1.5 to 4.5 for our potential error rate; remembering that ‘error’ rate in this case referred to the amount of meters found to have been tampered with. Our meter population was only known in the equation.

Taking all these factors into consideration, we used sampling tables to determine our sample size – 550 randomly-selected meters. (Source of tables: Sampling Manual for Auditors, The Institute of Internal Auditors). Those 550 meters represented 868,000 meters throughout APS’ 40,000 square miles of service territory, in 11 of Arizona’s 15 counties. 


Almost 12% of the sample size (64 meters) consisted of commercial accounts, which matched the actual percentage total. Just fewer than 35% of the sample meters were located outside the Phoenix metropolitan area, again matching the actual percentage total. The research study methodology was for APS electric servicemen to pull each of the selected meters, open the service panel, check voltage and make about 52 entries on a form specifically designed for this study.

We had also to determine what constituted theft or diversion, and we agreed that the laws of the state had made that determination for us. For outright theft or diversion, the standard ‘beyond a reasonable doubt’ would be used. For those instances where theft or diversion was suspected, the civil standard of ‘clear and convincing evidence’ was adopted. (An example of this would be excessive meter blade wear.) By setting these standards we also, by default, established a range into which our known and potential losses would fall.

The research study began on 3 April 2000, and all the data was received and compiled by revenue protection investigator Rebecca Allen by 30 June. The Audit Services Department then interpreted and translated the information, with the following results:

  • Definite meter tampering – 0.72%
  • Probable meter tampering – 1.00%
  • Actual loss dollars – $330,148
  • Actual loss revenue percentage -0.0215%
  • Probable loss dollars – $7,637,131
  • Probable loss revenue percentage -0.4965%
  • Total actual/probable loss dollars – $7,967,279
  • Total actual/probable loss revenue percentage – 0.5180%.

The data pointed to a much higher percentage loss at commercial accounts. Of the $7.9 million actual/probable loss, $5.1 million was attributed to commercial accounts. And, similar to the Canadian study, a large number of meter maintenance items were noted. Fully 6.5% of the meters in the study had some type of maintenance problem.

The value for APS is that we now know with a high degree of statistical certainty the percentage of meters that have been subjected to some form of tampering (1.72%), and the dollar value associated with actual loss ($7.9 million, 0.518% of revenues). The stratification of the study provides APS with information as to the type of theft that is occurring, and where. This allows for a more strategic allocation of assets. In an age of deregulation and the need to justify expenditures, this research study has already paid for itself.