20170215Blattner Rebuttal.pdf

Ronald L. Williams,ISB No. 3034
Williams Bradbury, P.C.
1015 W. Hays St.
Boise,ID 83702
Telephone : [phone]
Email: [e-mail]
Attorneys for Intermountain Gas Company
BEFORE TIIE IDAHO PUBLIC UTILITIES COMMISSION
IN THE MATTER OF THE APPLICATION OF
INTERMOUNTAIN GAS COMPANY FOR
THE AUTHORITY TO CHANGE ITS RATES
AND CHARGES FOR NATURAL GAS
SERVICE TO NATURAL GAS CUSTOMERS
IN THE STATE OF IDAHO
Case No. INT-G-16-02
REBUTTAL TESTIMONY OF LORI BLATTNER
FOR INTERMOUNTAIN GAS COMPANY
February 15,2017
)
)
)
)
)
)
)
1
2
J
4
5
6
7
8
9
l0
11
12
13
t4
15
t6
t7
l8
t9
20
2t
22
23
a.
A.
Please state your name, position and business address.
My name is Lori A. Blattner. I am a Regulatory Analyst with Intermountain Gas
Company ("Intermountain" or "Company"). My business address is 555 South
Cole Road, Boise, lD 83707.
Are you the same Lori Blattner that prepared and previously presented
prefiled direct testimony on behalf of Intermountain Gas Company in this
Case?
Yes
What is the purpose of your rebuttal testimony?
My rebuttal testimony will address the following areas:
1) Discuss the Weather Normalization case upon which the Company's weather
normalization process is based,
2) Discuss important differences between the Normal heating degree days used by
the Company in this proceeding and the Average weather year used by Staff
witness Morrison, and
3) Explain Staff s incorrect use of the Company's GS "subclasses".
WilI your rebuttal testimony discuss the statistical methods employed by Dr.
Morrison in the creation of his alternative weather normalization models?
No. The regression modeling is addressed by Dr. Philip Fry and Dr. Patrick
Shannon in Dr. Fry's rebuttal testimony.
lle at h e r Normalization C as e (U- I 0 3 4- 1 3 4 |
Please explain this case.
On July 30, 1986, the Company filed Case No. U-1034-134, an Application with
Blattner, Reb. 1
Intermountain Gas Company
a.
A.
a.
A.
a.
A.
a.
A.
I
2
J
4
5
6
7
8
9
10
1l
t2
l3
t4
15
l6
t7
18
r9
20
2t
22
23
a.
A.
the Idaho Public Utilities Commission requesting authority to utilize its proposed
weather normalization model for ratemaking purposes. In the Order from that
Case (OrderNo. 21048), the Commission "encourages IGC's efforts to improve
and refine its weather normalization methodology". The Company was
"authorized to implement its proposed weather normalization model for internal
forecasting and ratemaking purposes", with the Commission reserving decision on
its appropriateness until such time as it is presented in a ratemaking case for
consideration.
Has the Company been using this method since 1986?
Yes, the Company has been using the methodology outlined in the above
referenced Case since its authorization. The weather normalized consumption
levels resulting from this method have been used and incorporated in all of the
Company's PGA filings following 1986. These normalized therm sales have been
reviewed by the IPUC Staff during the annual PGA audits. Additionally, the
forecast usage for the months of March through October in the Company's
Integrated Resource Plans (IRP) have been forecast based on this method. Again,
these calculations have been reviewed by Staff as part of the Staff s audits of the
Company's IRP filings.
Who developed the weather normalization methodolory employed by the
Company?
The methodology was developed by Dr. John M. Kohlmeier, then a Partner with
Arthur Andersen & Co. Beginning on Page 1 of his Revised Direct Testimony in
Case U-1034-134 (see Exhibit 39) he outlines his professional credentials in the
Blattner, Reb. 2
Intermountain Gas Company
a.
A.
1 areas of quantitative methods where he specialized in modeling, planning and
forecasting.
a. Please explain the method.
A. The "Methodology Overview" found on Pages 2-3 of Revised Exhibit No. 1 from
Case U-1034-134 (see Exhibit 39) is as follows:
"To determine the degree towhich actual therms saleswere higher or lower
than normal as a result of actual weather, it is necessary to first quantify the
relationship betweenweather and sales. This quantification is achieved through the
use of multiple regression analysis at Intermountain Gas Company (IGC). Therm
usoge is statistically estimated as afunction ofweather and non-weather variables.
The relotionship between gas use and its main determinants is measured in three
regression equations, one which explains residential space heating-only sales, one
which explains residential space and water heating sales, ond one which explains
small commercial sales. All equations ore system wide.
To explain gas use, the regression equations use weather concepts, such
as heating degree days, and economic information, such as gas price. Combining
datafor all months of at least sixfiscal years in each regression allows enough
variationwithin the data to gain tnformationfrom more than one weother concept
andfrom economic variables. The summary statisttcs for the equations
demonstrate that the regressions are accurote in explaining monthly variations in
sales as is detailed in the methodologt descriptions. Each regression was tested
for autocoruelation and corue cted when ne ce s s ary.
Once the regression equations have been specified and estimated, it is the
Blattner, Reb. 3
Intermountain Gas Company
2
J
4
5
6
7
8
9
10
11
t2
13
t4
15
t6
t7
l8
t9
20
2t
22
23
I
2
3
4
5
6
7
8
9
l0
11
t2
13
t4
15
t6
17
l8
tea
20
2T A.
22
23
cofficients of the weather variables that are important to the weather adjustment
process. These cofficients measure the response of sales to changes in those
weather variables. For example, the cofficient of the heating degree day (base
65") variable in the space heating only, residential total system equation represents
the number of additional therms per customer that one additional heating degree
day (base 65") would cause. The cofficient of the heating degree day (base 45o)
voriable represents the number of additional therms per customer that one
additional heating degree day (base 45") would cause. Two weather variables are
used because an additional degree drop in temperature below 45" causes more
additional use of gas than a degree drop in temperature between 65" and 45". The
ffict of the HDD45 is additive to the effect of the HDD65. By multiplying these
cofficients by the dffirence between the normal and actual heating degree days
(base 65 " ) and heating degree days (base 45 " ), the dffirence between actual and
normal therms per customer is determined.
Note that the non-weather variables are used only to estimate the
regression equations and that only the cofficients of the weather variables are
used in the actual adjustment. Also note that the primary purpose of the models is
to adjust sales that have already occurued rather than to predict future sales. "
How are the Company's weather normalization models today the same or
different from the 1986 models?
In 1986, the models included a heating degree day term at both the 45 and 65
degree day level. As Dr. Kohlmeier stated in his testimony, the rationale for using
both the traditional 65 degree base as well as a second 45 degree base was to
Blattner, Reb. 4
Intermountain Gas Company
1
2
J
4
5
6
7
8
9
l0
11
t2
13
t4
15
t6
17 a.
l8
t9 A.
20
2l
22
23 0.
capture the fact that customers behave differently with respect to very cold
temperatures than they do to more moderate temperatures. He also included
monthly binary terms for October through April, and a Usage Trend variable. Each
year, the Company uses multiple regression analysis to develop regression
equations for each of its core market classes (RS-l, RS-2 and GS) on a total
company basis as was done in 1986. Over time, the variables have changed some.
With the warner weather we have experienced over the past decades, the HDD45
terms are no longer statistically significant, because we have not had enough
weather at low temperatures. As Dr. Fry discusses in his rebuttal testimony, the
Company also determined that monthly HDD65 coeffrcients better captured the
way customers respond differently to temperature at different times of the year.
For example, a degree day in June may not cause the customer to turn the fumace
back on, but an extra degree day in January will cause a fumace that is already on
to work a little harder. These variables incorporate what the HDD65 and the binary
variables together were measuring in the original models. A Usage Trend variable
is also included in the Company's 2016 models.
Wasn't the Usage Trend variable rejected by the Commission in the 1986
Case?
No. The Company had proposed an additional normalization adjustment to actual
usage based on the Usage Trend variable. This additional adjustment was
withdrawn by the Company. However, the Usage Trend variable properly
remained in the Weather Normalization models.
How did the proper construction of the weather variables ensure an accurate
Blattner, Reb. 5
Intermountain Gas Company
matching of gas usage and weather in the 1986 filing?
A. On page 4 of Dr. Kohlmeier's Revised Exhibit No. 1, he explained that "One of
the critical aspects in quantiffing the relationship between gas use and weather is
the correct matching of weather data with sales data". Heating degree day data
was collected from five NOAA weather stations: Caldwell, Boise, Idaho Falls,
Pocatello, and Twin Falls. Intermountain's service territory contains areas that can
have dramatically different temperatures at a given point in time. To incorporate
these weather differences, the weather variables needed to be constructed using
weather data weighted by the number of customers who experienced that same
local weather. Those divisions were represented by the Caldwell, Boise, Idaho
Falls, Pocatello and Twin Falls weather stations. Because residential and some
commercial customers are billed cyclically, usage data in a particular month
represents sales in the current month and previous month. To correctly match
weather and sales, the weather variables used in the analysis represent weighted
totals of the daily values of these variables for the month. Each day's weather
measure is weighted by the number of customers for a particular month who
experienced that weather.
a. Are the weather variables constructed in the same manner in the 2016 filing?
A. The weighting methodology is exactly the same in20l6 as it was in 1986. Over
the years, the Company has seen growth in both the Sun Valley and Rexburg areas.
To ensure that these areas were properly reflected in the weather data, the
Company began collecting weather from both Sun Valley and Rexburg. Thus the
weighting has been expanded from the original 5 weather stations to now include 7
Blattner, Reb. 6
Intermountain Gas Company
1
2
3Q.
4
5A.
6
7
8
9
l0
l1
t2
t3
t4 a.
15
t6 A.
t7
18
l9
20
2t
22
23
weather stations.
Companv Normal heatins desree davs vs. Staff Averase weather vear
Is StafPs Average Weather Year approach appropriate for ratemaking
purposes?
For the following reasons, the Company does not believe Dr. Morrison's Average
Weather Year should be used for ratemaking purposes:
1) The Average Weather Year used by Dr. Morrison was based onal4-year
average of heating degree days that ended 12 yearc before the test year,
2)The method used by Dr. Morrison is not a common industry practice for
calculating Normal HDDs nor is it a method supported by the National Oceanic
and Atmospheric Administration (NOAA),
3) The data chosen by Dr. Morrison was not weighted appropriately to reflect
Intermountain' s current customer base.
What definition of Normal heating degree days was used by the Company in
this Case?
As the Company discussed in pre-filed testimony (Blattner Di, page 4, line 3-12), a
Rolling 3O-year Normal was used by the Company. NOAA uses a 3O-year
definition ofNormal that is recalculated at the end of each decade. Many utilities,
including Intermountain, have tried to minimize the large "shock" that can come
with the every ten year update to Normal weather by utilizing a Rolling 30-year
Normal that is updated annually rather than at the end of each decade. As detailed
in their most recent Rate Case (AVU-G-l5-01), Avista also uses a 30-year rolling
average definition of Normal. Intermountain based its case upon Normal heating
Blattner, Reb. 7
Intermountain Gas Company
1
24.
3A.
4
5
6
7
8
9
10
1l
t2
l3
t4
l5
t6
t7
l8
t9
20
2l
22 a.
23
degree days from the 3O-year period ended December 2015.
How is this Normal weather used by Intermountain?
The Company collects heating degree day data from seven weather stations
including: Boise, Caldwell, Twin Falls, Sun Valley, Pocatello, Rexburg, and Idaho
Falls. The heating degree days for each day are srunmed and then averaged for the
30 year period. This process creates 365 days' worth of 30-year average heating
degree days for each of the seven weather stations. The daily data is then weighted
by the number of customers in each division represented by the weather stations
and by the billing cycles, as described in regards to the actual weather weighting
above. This is done separately for each rate class. The customer weights are based
on the most recently completed calendar year, in this case 2015. Because the
weather in Intermountain's service territory can be so diverse, this process allows
for an accurate matching between the weather data and the number of customers
that would experience that weather. The data was summed by revenue month and
became the Normal heating degree days used by the Company in the case. In the
actual months, actual heating degree days are compared to Normal heating degree
days and the monthly regression coefficient is applied to the difference to arrive at
the normalizing adjustment for the month for each rate class. In the forecast
months, rate class Normal heating degree days are used in the appropriate
regression equations to calculate the forecast monthly usage given Normal heating
degree days for each class (RS-I, RS-2, and GS-l).
Did Dr. Morrison use the same definition of Normal to develop his proposed
normalized usage?
Blattner, Reb. 8
Intermountain Gas Company
1A.
2
aJ
4
5
6
74.
8
9A.
l0
1l
t2
13
t4
l5
t6
t7
l8
te a.
20
2t A.
22
23
He did not. Instead, Doctor Morrison chose to experiment with an Average
Weather Year. As stated in Staff Response to Intermountain's Production Request
No. 13, "an average weather year is ayear consisting of average weather months".
For each month, Dr. Morrison used an average of a select series of RS-l monthly
Heating Degree Day (HDD65) values provided by the Company in its response to
StafPs Production Request No. 27.
What was the HDD65 data that was included in response to StaffProduction
Request No.27?
In response to Staff s Production Request No. 27, the Company provided a
download of the data that was stored inits eViews database. The Company
provided a dataset for RS-l, RS-2 and GS-I. That data only included actual data
from October 1989 through April20l5. In written discussions with Staffregarding
additional information related to Production Request No. 27 that was needed to
complete the review, Dr. Morrison stated, "You won't need to provide me
temperature data.I can download that myself if you provide me the list of
facilities from which the temperature information was obtained". Consequently,
the Company provided the list of weather stations, but no further heating degree
data.
In the Company's review of Dr. Morrison's worko how was the Average
Weather Year calculated?
Dr. Morrison used the HDD65 data from only the RS-l file (see the HDD tab of
Dr. Morrison's file included in response to Production Request No. 13, "PR #13-
3 6 Workpapers_Morrison_Normalization.xlsx"). The HDD65 s for each month
Blattner, Reb. 9
Intermountain Gas Company
were summed and that answer was divided by the number of months. In the
calculation of Average Weather, only data from October 1989 through May 2003
was used.
a. Was a reason given for the exclusion of the data from June 2003 through
April2015?
A. No reason was given. However, in my opinion, I believe it was an elror on Dr.
Morrison's part to exclude the most recent 12 years of weather data from the
analysis. Dr. Morrison included the recent l2 years of data in the calculation of his
regression equations, so it does not make logical sense to exclude those years from
the Normal HDD calculations.
a. Is the use of an Average Weather Year for normalization standard practice in
the natural gas industry?
A. I am not aware of another gas utility that uses an Average Weather Year for
normalization calculated in the manner used by Dr. Morrison. I am aware,
however, of the industry accepted use of the NOAA 30-year average or a Rolling
3O-year average. Some utilities use a slightly different time period, but none that I
am aware of use a fourteen year average that stops twelve years before the date the
case was filed.
a. Are there any additional problems with Dr. Morrison's Average Weather
Year approach?
A. Yes. The actual HDD65 data provided to Dr. Morrison in response to Production
Request No. 27 was customer and cycle weighted based on the actual customers
and cycles for each actual month. As the following chart illustrates, the percentage
Blattner, Reb. 10
Intermountain Gas Company
I
2
aJ
4
5
of customers living within the Company's warmer western region of the state has
increased compared with the percentage of customers living in the colder eastern
region. Thus, the weighting the Company used based on 2015 customers is much
different than the weighting would have been in the 1989 through 2003 data used
by Dr. Morrison.
Percent of Customers Living in the Western Region
66%
64/o
62%
60%
58%
s6%
54%
52%
1990 1995 2000 2005 2010 2016
a. What is the dollar impact to Test Year Base Rate Revenue as proposed by the
A.
Staff, caused by using Dr. Morrison's abbreviated and improperly weighted
Average Weather Year for normalization purposes?
Dr. Morrison has inflated his Normalized Test Year Base Rate Revenue by $3.6
million as outlined in Exhibit No. 40. The Average Weather Year calculated by
Dr. Morrison is colder than the Company's Normal heating degree days by 5Yo for
RS-l and GS-l and by 9% for RS-2. Using the colder Average Weather Year
heating degree days produces higher usage per customer when applied to Dr.
Morrison's models than does using Intermountain's Normal heating degree days
applied to those same models.
Blattner, Reb. 11
Intermountain Gas Company
6
7
8
9
10
l1
t2
13
t4
l5
l6
1Q.
)
J
4
5
6
7
8A.
9
10
1l
T2
l3
t4
15
t6
t7 a.
18 A.
t9
20
2t
22
23
In your opinion, is Dr. Morrison's proposal that the Company now set rates
based on Average Weather, instead of Normal heating degree days, and use a
l4-year average ofheating degree days that ended 12 years before the test
year, instead of using 30 years of weather history, a material change to this
Commission's sanctioned weather normalization methodolory for
Intermountain, as authorized in Commission Order No.21048 in Case No. U-
1034-134?
Absolutely. In my opinion, the proposed changes are so material as to render them
essentially a new methodology and a significant departure from the weather
normalization methodology authorized by this Commission for the Company in
Case No. U-1034-134. In making such a material deviation from prior
Commission approved practice, I believe the burden is on Dr. Morrison to prove
his methodology is better than the current methodology approved by the
Commission and used by other natural gas utilities. In fact, Dr. Morrison's
methodology is inferior, for the reasons outlined above.
GS "Subclasses"
Please define GS ttSubclasses".
GS "Subclasses" are an artificial construct based on the way the Company reports
billing data. The Company's billing data reports GS-10, GS-l1, GS-20 (until
August 2015), GS-60, and GS-12. The GS-60 designation tracks customers that
use natural gas to pump irrigation water, while GS-l2s are customers that use
natural gas for CNG vehicles. Although the remaining categories are tracked on
Intermountain's billing records, they do not represent customer groups with
Blattner, Reb. 12
Intermountain Gas Company
I
2
J
4
5
6
7
8
9
10
1l a.
t2
13 A.
l4
15
t6
t7
18
t9
20
2r a.
22 A.
23
homogeneous usage characteristics. They are instead internal billing designations.
GS-I0 denotes customers that are billed at the end of the calendar month while
GS-l1 denotes customers that are billed on a billing cycle basis. With the
implementation of the new CC&B customer information system in 2015, it was
determined that the GS-20 designation was indistinguishable from the GS-10
designation as both groups were billed on a calendar month basis. As the data
provided in response to Production Request Nos. 113 and 114 illustrates, the GS-
20 designation was rolled into the GS-10 designation beginning in August of 2015.
Other than the case of CNG, which has a separate pricing structure in the GS-l
tarifq these groups are all billed under the GS-l tariff.
Should any of the GS-l "subclasses" be treated differently for weather
normalization?
Yes. As Company witness Darrington's Exhibit No. 15, p. 8 illustrates, the
Company did not weather normalize the CNG (GS-12) or water pumper (GS-60)
sales in the actual months of the filing. The usage characteristics of both these
groups are very different from the characteristics of the rest of the GS-1 class.
Since GS-12 usage is relatively flat, the Company based the forecast months on
2015 usage. The forecast months for GS-60 were based on a three-year average.
Dr. Morrison agreed that weather normalization was not appropriate for these
customers.
What about the GS-10, GS-ll and GS-20 designations?
While these categories are listed on lntermountain's billing records, they do not
represent customer groups with homogeneous usage characteristics. They are
Blattner, Reb. 13
Intermountain Gas Company
1
2
J
4
5
64.
7A.
8
9
10
11
t2
13
t4 a.
15
16 A.
t7
18
t9
20
2l
22
23
instead intemal billing designations. As noted above, GS-20 is no longer a
designation tracked in the system. It is therefore inappropriate to segregate these
"subclasses" for weather normalization. The Company uses the aggregated data
from these designations upon which to base its total GS-l regression analysis and
weather weighting.
Did Dr. Morrison also use aggregated GS-l data in his analysis?
No. Dr. Morrison decided to use the detailed billing data to segregate the GS-l
class and build three separate weather normalization equations. As Dr. Morrison
outlined in his Direct Testimony (page 3, lines l5-16), "my analysis of the
Company's GS-l subclasses is more detailed than the Company's". He created
separate equations for GS-10, GS-11 and GS-20. As noted previously, the GS-20
designation was rolled into the GS-10 designation in 2015 and does not exist
separately today.
Do the consumption patterns of these subclasses differ sufficiently to warrant
separate treatment?
No. Since the subclasses are only designations made by the Company to identifu
the meter reading schedule of the customer, there is nothing distinguishable about
the customers. Breaking the class into artificial groups does not enhance the
overall accuracy of the equations. In fact, as Dr. Fry's rebutial testimony points
out, Dr. Morrison's equations for GS-l I and GS-20 include variables that are not
statistically significant. In contrast, the Company's single model based upon the
entire GS-l class provides a model with sound statistics as Dr. Fry demonstrates in
his rebuttal testimony.
Blattner, Reb. 14
Intermountain Gas Company
I
2
a.
A.
Does this conclude your Rebuttal testimony?
Yes.
Blattner, Reb. 15
Intermountain Gas Company