Methodology of the Teletruth-UCAN Phone Bill Survey and Report.
This document is designed to examine the methodology and sampling issues used by Teletruth as well as examine the other flawed methodologies used by the FCC and suggested by our critics.
A number of critics of our study have complained about the overall wireless cost-per-minute statistic. The method proposed by some critics is the same methodology used by the FCC. --- all costs divided by all minutes. The FCC recent report in January 2009, claims that the cost of a one minute wireless call was $.06 in 2007. The FCC method, however does not use actual phone bills nor does it differentiate low, medium or high-volume customers. The FCC's analysis is always skewed towards heavy users. Unfortunately, it can be used politically to show that wireless prices are in decline.
We will come back to address the FCC's data.
Another false method of examining cost per minute is to use the "median", which is an examination based on counting the 50% marker of the surveyed customers. Our study does show that for all users, about 50% are paying under $.25 and the second half is paying over $.25. However, if you 'average' each of the sections, the first half averages $.15 a minute, while the second half averages $4.94 a minute. Sample Using Low, Medium and High Volumes Break Outs.
In our study we created a matrix of user groups, categorized by minutes of use. 18% of our sample were ‘low volume’ customers using under 10 minutes of airtime. 14% of our sample were high-volume customers using over 1000 minutes. (UCAN members.)
The minutes of use by category are not under contention. The chart below indicates the groups are fairly close. Furthermore, our 1993 and 1995 nationwide surveys showed this breakout of high-medium and low volume usage was almost identical  The FCC Method Skews the Data Toward High Volume Users.
The next step is to chart the minutes of use by category. If we assumed 100 phone bills fitting the ranges of high-medium low, we would have 14 customers (14%) with over 1000 minutes, as well as 18 customers (18%) with 10 minutes or less, etc. This next chart shows that the two highest user categories, 300-1000 minutes and over 1000 minutes dwarf the other categories in total minutes.
80-20 Rule Applies to Telecom Too:
What is occurring is that the high-volume customers are taking over all other categories, skewing the numbers towards these two groups. In the following chart we see that the 1000+ customers who only represent 14% of the users, now have 49% of minutes used. Combining the next largest group, 300-1000 minutes, -- 89% of the minutes are accounted for. --- in this case it's 89-40 rule... 40% of the customers make 89% of the calls.  What this does is negate the other users. For example, many customers in the under 10 or less minute group have plans costing $20.00-$40.00, and these customers have ‘costs per minute’ that are $10.00-$20.00. Based on our survey, it is clear that these are not "savvy" consumers. Even though many argued these consumers do not exist or failed to take responsibility, our survey results do not agree. Counting Accounts: Subscriber-based Analysis. The data we presented for 'cost per minute' and the difference with other methods can be seen in the following exhibit. Below are the first 5 Verizon phone bills in our database. For example, phone bill Number 4010, has a total cost of $64.76, used 48 minutes, had a 500 call allowance, and the average cost was $1.35 per minute. The next bill had a total cost of $18.57, made only one minute of calls and the cost per minute is $18.57.
When we average the costs for these 5 bills, average bill was $45.26, with an average of 103 minutes of calls, (about 1/3 of the allowance) and the average cost per minute, when calculated by accounts was $5.78. ‘Combining all users”.
There is another analysis that some would consider – combining all users and all minutes, i.e., $45.26 divided by 103 equals $.44 a minute, not $5.78. The problem with that method is that it still weights the numbers and doesn’t examine the information ‘per-account’, The FCC's Data on Wireless Calls is Seriously Flawed.
In January 2009, the FCC released "13th Annual CMRS Competition Report" It made the following claim. "On average U.S. mobile subscribers paid about $0.06 per minute for mobile voice calls in December 2007." The FCC's data on wireless service is essentially an artificially created statistic that has no bearing on the reality found on phone bills. We've commented extensively about the FCC's data as the problem runs throughout their statistics, even though they are used for the creation of regulation and the information for Congress.
The FCC Does Not Use Phone Bills.
The Agency relies on estimates and revenue data, complied by the companies and their association, the CTIA. "Some analysts believe average revenue per minute (“RPM”) is a good proxy for mobile pricing. This is calculated by dividing a provider’s estimate of average monthly revenue per subscriber (often referred to as average revenue per unit, or “ARPU”) by its estimate of MOUs (minutes of use), yielding the RPM that the provider is receiving. Using estimates of industry-wide ARPU and MOUs from CTIA’s survey, (the wireless association) we estimate that RPM was $.06 in December of 2007, which is a decrease of one percent from December of 2006." Does this calculation include the taxes or surcharges applied, including made up fees, such as the 'cost recovery fee" or the plan fees or the various taxes and surcharges, some of which are pass-throughs paid by customers? We note that no one in our survey was paying $.06 a minute, so it can't be the average -- that would require a large group to have $.01 or less cost per minute. The FCC's Data Does Not Break Down the Actual User Groups. As we've described, when you aggregate all of the calls the high volume calls drown out the data pertaining to the majority of calls.
A number of critics believe that the cost per minute should be done using a median analysis -- As defined in Wikipedia http://en.wikipedia.org/wiki/Median "A median is described as the number separating the higher half of a sample, a population, or a probability distribution, from the lower half. The median of a finite list of numbers can be found by arranging all the observations from lowest value to highest value and picking the middle one." Sounds reasonable but the mathematics shows that it does not represent the low or even middle volume customers. Below is the reason. While 51% of accounts are paying under $.25, the other group is not simply paying over $.25; 28% are paying over $1.00 a minute.
Also, the majority of 2 line accounts -- 75% --- are paying under $.25%, while only 31% of one line accounts are paying under $.25. thus, the analysis should be separated for low and mid-volume users. as well as one and 2 line accounts. Notice the huge difference between one and 2 line costs per minute.
If we use the median analysis, 51% are paying under $.25, with an average of $.15; conversely, the second group is paying a whopping $4.94 a minute .
Sampling Issues The goal of the survey was to see what changes had occurred in telephone and broadband service since our previous survey with UCAN conducted in 2004, which collected over 200 wireline and over 125 wireless bills. We believe the sample is a solid representation of San Diego CA customers and may have implications throughout the US based on previous work done by New Networks Institute, LTC Consulting and Teletruth as wireless and long distance plans are created as national plans, that do not change dramatically state by state. We agree that more survey data throughout the US and even California is necessary to determine what variances may occur. Collecting Bills UCAN announced the survey in a monthly newsletter in Spring, 2008 and it generated over 700 phone bills, 572 wireline, local, long distance, cable-with-phone service, and broadband bills and 134 wireless bills, for all carriers. Most of these phone bills were inputted into a proprietary database that could compare the 2008 phone bill data to the phone bills collected in 2004.( In the case of wireless, we inputted 117 bills as the others were either part of a package or was missing pages.) The report data was generated using specific analysis tools built into the database. Additionally, Teletruth called over 90 customers from the 2008 survey, of which 20 agreed to a full interview. Does the UCAN sample skew the data or does it represent San Diego, much less the US? UCAN is a consumer advocacy group and its members get information about phone, broadband and other utility bills. One would expect the sample to represent 'better educated' customers who should have a better handle on costs. Are UCAN members 'atypical users' - do they make more minutes? Less minutes? What we find startling is that if this sample represents an 'educated consumer', the rest of the population could, in fact, have worse wireless and wireline bills. UCAN members might trend toward older users than the general population, which could lower usage. However, we can't tell if the 'younger' UCAN members were less likely to be part of the survey - with 700+ bills, we don't believe it seriously skewed the data. Sample Size Most surveys use 1000-2000 interviews to do a nationwide sample, though it varies by research/polling needs. The country has an estimated 300 million people (US census 2007). San Diego had 1.27 million in 2007, so the San Diego 117 cell phone bills should be enough for a strong sample of a city. Also, since over 80% were AT&T and Verizon, the survey has a good representation of the 2 largest carriers. National Implications We can make a number of statements about what we would expect to find on a national basis because of our previous surveys and the fact that the wireless services are being offered with national plans. Long distance service is also offered with national plans, for example by AT&T. Also, in our 1993 report, "Consumer Attitudes and Perceptions Toward Cable and Phone Companies", Probe Research, publisher) with 1000 consumers, we found that …"throughout the questions asked, the variance between answers for any region or any state was nominal and statistically insignificant." This included questions about usage of local and long distance, 'active-vs-inactive users' and even knowledge about the cost of directory assistance in their state. |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
