Analysis of Potential Demand for the Extension and Expansion of

Analysis of Potential Demand
for the Extension and Expansion of
Natural Gas Distribution Infrastructure
in Pennsylvania
A Report in Response to Senate Resolution 29
November 2013
Analysis of Potential Demand for the Extension and Expansion of
Natural Gas Distribution Infrastructure in Pennsylvania
A Report in Response to Senate Resolution 29
Richard Ready, Ph.D., Principal Investigator
The Center for Rural Pennsylvania is a bipartisan, bicameral legislative agency that serves as a resource
for rural policy within the Pennsylvania General Assembly. It was created in 1987 under Act 16, the Rural
Revitalization Act, to promote and sustain the vitality of Pennsylvania’s rural and small communities.
Information contained in this report does not necessarily reflect the views of individual board members or
the Center for Rural Pennsylvania. For more information, contact the Center for Rural Pennsylvania, 625
Forster St., Room 902, Harrisburg, PA 17120, telephone (717) 787-9555, email: [email protected],
www.rural.palegislature.us.
Preface
In March 2013, the Senate adopted Senate Resolution 29, directing the Center for Rural Pennsylvania to
study the potential for the increased extension of natural gas distribution infrastructure by Pennsylvania’s
natural gas public utilities to unserved and under-served areas.
Specifically, the Center was charged with studying the residential, commercial and industrial extension
of natural gas distribution infrastructure by collecting and analyzing information on the:
• estimated demand for natural gas service in unserved and under-served areas of the commonwealth;
• estimated price consumers are willing to pay for access or conversion to natural gas service;
• regional differences in consumer demand and willingness to pay for natural gas service; and
• other relevant economic information on the costs and benefits to expand natural gas distribution infrastructure.
To consider residential extension, the Center worked with researchers to conduct a telephone survey of
Pennsylvania households and developed a demographic and socioeconomic profile of Pennsylvania communities.
The household survey included more than 1,000 Pennsylvanians from four regions of the state. These
regions encompassed both rural and urban areas and were selected to provide geographic and demographic
diversity to the research.
The survey revealed several important findings including the following: the study respondents were well
informed about the relative operating costs of different heating systems; there may be a large potential
pool of households interested in connecting to natural gas to save money on heating; and the probably of
households converting to natural gas service increases as the upfront costs decrease and as the payback
time decreases.
To complement these and the other research findings, the Center for Rural Pennsylvania developed a
demographic and socioeconomic profile of Pennsylvania communities in which natural gas distribution
infrastructure is and is not available. The profile is in Appendix 1. Additional data and maps are provided
in Appendix 2 to provide insight for potential industrial/commercial service expansion.
The Center thanks Dr. Richard Ready, the lead researcher who developed the survey instrument and
analyzed the findings, for his work. The Center also thanks Mr. Berwood Yost, director of the Center for
Opinion Research at Franklin and Marshall College, and his staff, for administering the telephone survey.
Barry L. Denk
Director
Executive Summary
A telephone survey was conducted to determine the proportion of Pennsylvania homeowners who would
connect to natural gas distribution lines if they were made available to save on home heating costs, and how
much those homeowners would be willing to pay for a connection. The survey included 1,020 households
in 15 counties.
The survey showed that most Pennsylvania homeowners are well informed about the cost-savings potential of natural gas heat, and that most Pennsylvania homes that are not now connected to natural gas could
be converted to natural gas as a primary heating source.
In the survey, respondents were presented with a series of hypothetical connection scenarios. In each
scenario, respondents were asked whether they would connect to natural gas and use it to heat their home
if given the opportunity. The cost of connecting to the natural gas line and installing the necessary equipment in the home and the projected annual savings in heating costs from doing so varied across scenarios.
The survey results showed that the proportion of households who would choose to connect decreases as the
upfront costs increase and increases as the projected annual savings increases. The results also showed that
homeowners in the North Central region of the state are more likely to connect than homeowners located in
the South Central and South Eastern parts of the state.
A statistical model was estimated that predicts the proportion of households that would connect to a new
natural gas distribution line. For a typical household facing upfront costs of $6,000 and annual savings of
$1,000, the model predicts that 17-25 percent of households in the South Central and South Eastern portions of the state would connect while 26-32 percent of households in the North Central part of the state
would connect. The model further predicts that half or more of Pennsylvania households would not connect
to a natural gas line regardless of the upfront costs or payback period on the investment. Reluctance to connect is related to three attitudes/perceptions: worry about potential future increases in gas prices, the hassle
of installing a new supply line and heating equipment, and an inability to afford upfront costs.
When considering construction of a new distribution line, natural gas distribution companies must be able
to project how many households will choose to connect to the new line. The information generated in this
study can help in making that assessment.
Contents
Study Purpose.....................................................................................................................................................6
Interview Methodology.......................................................................................................................................6
Results.................................................................................................................................................................7
Demographics of Survey Respondents............................................................................................................7
Identifying Homes that Might Connect to Natural Gas...................................................................................8
Homes that Have Natural Gas but Do Not Use It for Primary Heat................................................................9
Perceptions Regarding Home Heating Costs...................................................................................................9
Natural Gas Connection Scenarios..................................................................................................................9
Factors Influencing Connection Decisions....................................................................................................10
Analysis of Willingness to Pay Responses....................................................................................................10
Modeling Factors Influencing Connection Decisions...................................................................................11
Predicting Connection Rates.........................................................................................................................13
Conclusions.......................................................................................................................................................15
References.........................................................................................................................................................15
Appendix 1: Demographic and Socioeconomic Profile of Pennsylvania Communities..................................16
Appendix 2: Business and Industry Profile by County.....................................................................................25
Appendix 3: Senate Resolution 29....................................................................................................................28
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
5
Study Purpose
According to production data from the U.S. Energy
Information Agency, natural gas production in Pennsylvania averaged over 6 billion cubic feet per day in
2012, a 10-fold increase over production levels 5 years
prior. Over that same period, natural gas retail prices
fell by one quarter. Lower natural gas prices have benefited Pennsylvania residents who use natural gas to heat
their homes through lower home energy bills. Natural
gas is currently one of the cheapest ways to heat a
home. However, many Pennsylvania homes do not have
access to natural gas distribution infrastructure.
Senate Bills 738 and 739 aim to increase access
to natural gas distribution infrastructure for homes,
schools, hospitals and small businesses. The benefit
from investments in new distribution infrastructure
aimed at serving homes will depend on how many
newly-served homes connect to the new infrastructure,
and what appliances they choose to convert to natural
gas. These homeowner decisions will depend on the
cost of connection and appliance conversion and the
perceived cost advantage from adopting natural gas.
Further, some homeowners may choose to hook up only
when current appliances need to be replaced.
This research determined the proportion of homes
that would connect to natural gas if it were made available to them, how that proportion would vary at differ-
ent connection costs (i.e. willingness to pay to connect),
and what factors would discourage homeowners from
connecting to newly-available service (barriers to connection), such as structural limitations in their home,
concern over the disruption and inconvenience associated with connection to natural gas lines, and concern
over natural gas price volatility.
This was done using a survey of Pennsylvania
homeowners. The centerpiece of the survey was a set of
questions that measured how much homeowners would
be willing to pay (WTP) to connect to natural gas
service and convert to natural gas heat for their home.
Specifically, homeowners were presented with different
scenarios regarding the upfront costs to connect/convert to natural gas and the anticipated annual savings
that would result, and were asked whether they would
or would not connect/convert under those scenarios.
Additional questions revealed non-financial factors that
influence the connection decision.
Interview Methodology
The telephone survey, conducted from July 15 to
August 9, 2013, included interviews with 1,020 homeowners in four regions of the state. The Natural Gas
Infrastructure Survey was administered by the Center
for Opinion Research at Franklin and Marshall College.
To focus survey efforts on households that do not
currently have natural gas
Table 1. Counties included in the telephone survey
service, the research focused on 15 counties in four
regions with low natural gas
penetration rates. Data on
the percent of households
using natural gas for heat in
the study counties are shown
in Table 1. While the four
regions vary geographically
and demographically, the
survey results are representative of homeowners in the
selected regions and are not
intended to represent all
households in Pennsylvania.
Households were randomly
selected from each region
using traditional randomdigit-dialing methods. To
minimize sample selection
bias, both landline and cell
phone sample frames were
a
Includes owner-occupied and rental units. Survey only includes owner-occupied units. Source: American
used. Eleven percent of the
Community Survey, U.S. Census, 2010.
6
The Center for Rural Pennsylvania
Map 1. Study regions and counties
households interviewed were cell-phone-only households. An adult household member who could respond
about the home’s heating system was asked to complete the survey whenever an eligible household was
identified by the interviewing staff. Householders were
eligible to participate when they owned their homes and
those homes had their own heating systems.
The telephone interviews were conducted using
a computer assisted telephone interviewing (CATI)
system by the Center for Opinion Research. The CATI
system controls the flow and ensures the logic of the
questionnaire, catalogues data to administer and manage the interviewing process, and stores the data for
cleaning, coding, and processing. Table 1 shows the
number of completed interviews, by region/county.
Map 1 shows the study regions and counties.
All sample surveys have potential sources of bias or
error and this survey is no exception. Generally speaking, two sources of bias concern researchers most:
non-response bias and response errors. Non-response
bias is created when selected participants either choose
not to participate in the survey or are unavailable for
interviewing. In the current survey, the cooperation
rate was 92 percent. The cooperation rate represents
the proportion of all cases interviewed of all eligible
units ever contacted. The response rate is 57 percent
(calculated using the American Association of Public
Opinion Research response rate 3 calculation). Information on response and cooperation rates is not widely
published, so it is difficult to contextualize these rates.
A recent published analysis of response rates (Curtin,
Presser, and Singer 2005) using Michigan’s Survey of
Consumer Attitudes showed a decline in response rates
of about 1 percent per year between 1979 and 2003,
with the 2003 survey’s response rate being 48 percent.
Response errors are the product of the question and
answer process. Surveys that rely on self-reported behaviors are susceptible to biases related to the way respondents process and respond to survey questions. The
inability to accurately recall events, to accurately recall
behaviors that occurred within a specific time frame,
or to provide honest responses are each of concern to
researchers who collect this type of data.
Results
Demographics of Survey Respondents
The majority of survey respondents (86 percent) live
in single family, detached homes. Eight percent live
in attached homes (duplexes, townhomes, etc.), while 6
percent live in mobile homes. By design, all respondents
owned their own home, and paid for their own heat.
The average respondent lives in a home that is 55
years old, with a heating system that is 11.7 years
old. Fifty-nine percent of respondents had replaced or
changed their heating system since moving into the
house, indicating that most respondents were familiar
with making heating system investments. Respondents
were told that the average home in Pennsylvania is
around 1,800 to 2,000 square feet in size. Forty-nine
percent of respondents indicated that their home was
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
7
about average in size, 29 percent said their home was
larger than average, and 20 percent said their home
was smaller than average (2 percent did not know).
Fifty-three percent of respondents had invested in extra
insulation or weather stripping for their home and 40
percent said that their home was better insulated than
average. Only 14 percent said their home was less well
insulated than average, which suggests that residents
may tend to overestimate their level of insulation.
Thirty-six percent of respondents have a setback or programmable thermostat, and 16 percent had an energy
audit conducted on their home.
Eighty-four percent of respondents said it was likely
or very likely they would still be living in their house
in 5 years, while 67 percent said it was likely or very
likely they would be still living there in 10 years, indicating that most Pennsylvania households plan to stay
in their home long enough to benefit from a cost-saving
investment in a new heating system.
The average respondent lives in a household with 2.7
members and is 56.7 years old. Fifty-five percent of
survey respondents are female, 43 percent of respondents have a college degree and 72 percent are married.
Forty-two percent of respondents work full-time, 12
percent work part-time, 31 percent are retired, 7 percent
are homemakers, 3 percent have a disability, 2 percent
are unemployed, and 1 percent are going to school.
The average reported household income was $68,500,
though 26 percent of respondents chose not to answer
the income questions in the survey.
Compared to Pennsylvania as a whole, the survey
sample is slightly more female (Pennsylvania proportion is 51.2 percent), lives in a slightly larger household
(state average is 2.47), is more likely to have completed
college (state proportion is 26.7 percent), has higher income (state median household income is $51,700), and
is less likely to be unemployed (state unemployment
rate was 7.8 percent in July 2013). These differences do
not necessarily mean that the sample is biased relative
to the target population for the survey. The population of interest for this study is households who own
their own home. Homeowners are likely to have higher
incomes, higher rates of college education, and lower
unemployment rates than households who do not own
their own home.
Identifying Homes that Might Connect to
Natural Gas
The following types of households were excluded
from the analysis, because they were assumed to be unlikely to connect to natural gas service for their primary
heating:
• households that already have natural gas service;
• households that use a geothermal heat pump for
their primary heating (heating with a geothermal
heat pump is similar in cost to heating with natural
gas, so such households would have little incentive
to switch); and
• households that do not currently have ductwork
or pipes that can distribute heat from a furnace or
boiler and that cannot add ductwork or pipes because of structural limitations of the home.
Table 2. Initial sample sizes and exclusion rules
8
The Center for Rural Pennsylvania
Table 2 lists the number of respondents in
each region that were excluded due to each
of the three requirements, and the resulting
samples of households that could potentially
connect to natural gas for their heating needs.
The proportion of households who do not
currently have a natural gas connection but
that might want to connect to natural gas for
their heating was highest in the South Eastern region since very few homes in the South
Eastern region have structural limitations that
would prevent them from installing ducts or
pipes for a new natural gas heating system.
Table 3. Which heat sources respondents thought
were most and least expensive
Homes that Have Natural Gas but Do Not Use
It for Primary Heat
Across the four regions, the sample included 48
homes that currently have a connection to a natural
gas distribution line, but that do not use natural gas as
their primary heating source. These households are of
interest because it would be expected that they could
save money by switching to natural gas as their primary
heating source, but they have not done so. Additional
questions were asked of these respondents to determine
why they have not already switched to natural gas heat.
Of the 48 respondents, 18 use natural gas as backup
heat. All 48 respondents were asked whether they
thought it was likely that they would convert to natural
gas as their primary heat source within the next 5 years.
Twelve respondents (25 percent) indicated that it was
likely or very likely that they would. Of those, the most
common reason was cost and efficiency (5 respondents), followed by ease of conversion due to already
having service (3 respondents), convenience of natural
gas (2 respondents), and the need to replace an existing system (2 respondents). Thirty-three respondents
(69 percent) indicated that is unlikely or very unlikely
that they would convert to natural gas as their primary
heating source. The most common reason was that the
homeowner was satisfied with their current system and/
or the current system did not need to be replaced (14
respondents), followed by the cost of conversion (7
respondents), the respondent does not like natural gas
as a heating source (4 respondents), and the respondent
is not familiar with natural gas (2 respondents).
Perceptions Regarding Home Heating Costs
The 552 respondents identified as potential natural
gas customers were asked which heat source was usually the most expensive to heat a home and which was
usually the least expensive. The responses are sum-
marized in Table 3. Most respondents correctly identified natural gas and geothermal heat pumps as the least
expensive heat sources, and fuel oil and electric (resistance) heat as the most expensive.
Natural Gas Connection Scenarios
Respondents identified as potential new natural gas
customers were then told to imagine that natural gas
service was made available to them, and asked questions to measure how much they would be willing
to pay to convert to natural gas as their primary heat
source. Respondents were told that, if they were to convert to natural gas as their primary heating source, they
would likely save money on their annual heating bills,
but that converting to natural gas would require spending money upfront to connect their house to a natural
gas distribution line and to purchase and install necessary equipment in the home.
Each respondent was given three scenarios. In each
scenario, the respondent was told how much he or she
would have to pay in upfront costs to convert to natural
gas, and how much he or she would save each year in
heating costs compared to his/her current heating system. The payback time (in years) for a scenario is then
the upfront costs divided by the annual savings. A total
of nine different scenarios were used in the survey (See
Table 4 on Page 10). The ranges of upfront costs and
annual savings were chosen to represent the range of
costs and savings that Pennsylvania households would
likely experience if they chose to connect and convert
an existing home to natural gas heat. The midpoint
values ($6,000 upfront cost and $1,000 annual savings) were chosen based on publicly available estimates
from natural gas companies, media outlets, and the
U.S. Energy Information Administration. Higher and
lower upfront costs were chosen to represent homes
with above-average and below-average connection and
conversion costs. For example, a home located close to
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
9
Table 4. Scenarios used in
WTP questions
WTP questions, 288 (52.2 percent) answered “No” or
“Don’t Know” to all three scenarios presented. These
respondents were asked two follow-up questions to
help identify reasons why households might not be
interested in natural gas heat. The results of those two
questions are shown in Table 6. The follow-up questions show that the inability to pay upfront connection
and conversion costs are more common reasons why
respondents would not connect to natural gas than
technical difficulties in retrofitting the home’s heating
system.
Analysis of Willingness to Pay Responses
The validity of using hypothetical questions to
measure WTP for natural gas service depends on the
assumption that survey respondents answer the hypothetical questions the same way that they would make
actual connection decisions. Much research has been
conducted to compare answers to hypothetical WTP
questions against actual behavior, in a variety of WTP
contexts. This research has shown that survey respondents tend to overstate the likelihood that they would
pay the amount of money posed in a WTP question.
This tendency is called “hypothetical bias.”
a distribution line that already has a forced air system
might face lower than average connection and conversion costs, while a home that does not have ducts or
pipes already installed could face higher than average
conversion costs. Annual savings would be higher or
lower than average based on heating needs and the
efficiency of the current heating system. A smaller
home that currently heats with an air-source heat pump
would face lower than average annual savings while
a large home that currently heats with fuel oil would
face higher than average annual savings. High,
average and low values of upfront costs and
Table 5. Agree/Disagree questions on factors
annual savings were combined into nine difinfluencing connection decision
ferent combinations to generate information
on the independent effects of each.
Each respondent was presented three of the
nine scenarios, chosen randomly. For each
scenario, the respondent was asked whether
he/she would or would not convert to natural
gas as his/her primary heating source.
Factors Influencing Connection
Decisions
The 552 respondents who were asked the
WTP questions were asked three agree/disagree questions regarding factors that might
influence whether they connect. The responses
to these three questions are shown in Table 5.
A strong majority agreed that connecting to
natural gas for heat would be a “big hassle”
and that they worried whether gas prices might
rise in the future. Fewer than half of respondents indicated that they would only switch
to natural gas heat when their current system
needed to be repaired or replaced.
Of the 552 respondents who were asked the
10
The Center for Rural Pennsylvania
Table 6. Agree/Disagree questions why respondent
would not connect to natural gas
of 7 or 8 generated recoded hypothetical
responses that best matched actual behavior.
For this study, all WTP responses are
recoded using both a threshold of 7 and a
threshold of 8. Using a threshold of 7 results
in larger predicted connection rates than
using a threshold of 8. Using both thresholds generates a range of predictions for
adoption rates. Specifically, the recoding
was conducted as follows for each WTP
response. All respondents who said “no” or
“do not know” to the WTP question were
treated as negative responses. Respondents
who said “yes” to the WTP question but
who then gave a certainty rating less than
7 (or 8) were treated as negative responses.
Only respondents who said “yes” to the
WTP question and then gave a certainty
rating of 7 (8) or higher were treated as
positive responses.
Figure 1 on Page 12 shows how the percent of posiFor each of the three scenarios presented, after answering the “would you convert” question, respondents tive responses (after recoding) varies with upfront costs
and payback period, using a certainty threshold of 7.
were asked how sure they were about their answer, on
As a general rule, respondents are more likely to give
a scale from 1 (very unsure) to 10 (completely sure).
a positive response if the upfront costs are lower and
A common approach to remove hypothetical bias from
if the payback period is shorter. For the most attractive
WTP responses is to recode positive responses where
scenario, with $3,000 upfront costs and a 3-year paythe respondent states a low level of certainty. Several
back period, 40 percent of respondents gave a positive
studies have supported this approach. For example,
Ethier et al. (2000) compared hypothetical responses to response (using a certainty threshold of 7), while for
the least attractive scenario, with $10,000 upfront costs
a WTP question for a green energy program to actual
signup rates for the program. They found that recoding and a 20-year payback period, 13 percent of responall positive hypothetical responses with a certainty level dents gave a positive response. Figure 2 on Page 12
shows the same information using a certainty threshold
less than 7 (on a 10-point scale) to negative responses
of 8. Using a higher certainty threshold results in lower
generated recoded hypothetical behavior that matched
actual behavior. Other studies have found similar results proportions of positive responses for all up-front-cost/
payback-period combinations. With the higher certainty
for a range of contexts. Table 7 lists eight studies that
threshold, the proportion of positive responses varied
used a 10-point certainty scale to calibrate hypothetical WTP responses. For each study, the certainty cutoff from 7 percent to 30 percent.
that best matched actual behavior is shown. In all cases
Modeling Factors Influencing Connection
except one, the studies found that a certainty threshold
Table 7. Review of studies using certainty
scales to calibrate WTP responses
Decisions
To systematically predict connection rates for any
combination of upfront costs and payback period, and
to explore other factors that influence connection rates,
a statistical model was estimated. A logistic regression
model was estimated that predicts the probability of a
positive response according to the formula
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
11
Figure 1. Percent positive response
using a certainty threshold of 7
Table 8. As expected from Figures 1 and 2, the probability of a
positive response is lower if either
upfront costs are higher or the
payback period is longer. Both
results were statistically significant at the 1 percent level for both
certainty thresholds.
Respondents with higher
incomes (measured in $1,000)
were significantly more likely
to connect. A dummy variable
indicating that the respondent
did not report his/her income in
the survey was not statistically
significant. Respondents who are
likely to move within 10 years
were significantly less likely to
connect, but only in the regression using a certainty threshold
of 8. This result makes sense,
since households who plan to
move soon would enjoy the anFigure 2. Percent positive response
nual savings for fewer years. The
using a certainty threshold of 8
probability of connecting was
significantly higher for female
respondents, lower for married
respondents, and lower for older
respondents. Education (whether
the respondent was a college
graduate), employment status
(whether the respondent had a
full-time job) and household size
were not significant predictors.
Respondents who agreed with
the statement “I am worried that
natural gas prices could go up in
the future if I switch to natural
gas” were less likely to connect,
as were respondents who agreed
with the statement “Switching to
natural gas would be a big hassle
because a new supply line and
equipment are needed,” suggesting
that
these
two
concerns
are important impediments
where X1, X2,… are factors that influence the probabilto
connection.
In
contrast,
respondents
who agreed with
ity of a positive response (upfront cost, payback period,
the statement “I would only switch to natural gas if and
geographic region, characteristics of the respondent,
when my current heating system needed to be replaced
etc.) and β0, β1, β2,… are model parameters to be estior repaired” were not significantly less likely to conmated in the regression.
A model was estimated that explored 25 different fac- nect.
Several physical characteristics of the home that
tors that might influence the connection decision. The
would be correlated with heating costs turned out not to
estimated parameters from this model are presented in
12
The Center for Rural Pennsylvania
Table 8. Logistic regression results
placed their heating system at
some point was not a significant predictor. Curiously, respondents who had an energy
audit performed were significantly less likely to give a
positive response, a result that
is somewhat counterintuitive.
Respondents who stated that
their heating bills were lower
than the average household’s
heating bill were less likely to
give a positive response.
Respondents living in the
South Eastern region, the
South Central region, and
Cumberland County were
significantly less likely to
connect than respondents in
the North Central region. It
is interesting to point out that
the North Central region is
the only study region located
in areas with active drilling
for Marcellus Shale natural
gas. Additional regressions
showed that the probability
of connection did not differ
significantly among the South
Eastern region, the South
* - statistically significant at the 5 percent level; ** - statistically significant at the 1 percent level
Central region, and Cumberland County. The population density of the respondent’s
be significant predictors of connection. These included
zip code was not statistically associated with the probage of home, age of the heating system, whether the
ability of connection.
home was better insulated than average, and whether
the home was larger than average.
Predicting Connection Rates
Respondents who had taken actions in the past to
While the regression results shown in Table 8 are
reduce heating costs, specifically adding insulation
interesting in that they reveal who is more likely or less
to their home or installing a setback thermostat, were
likely to connect to natural gas, they cannot be used
more likely to connect. Whether respondents had refor predictive purposes for
Table 9. Logistic regression results for predictive model
households that were not included in the survey. To make
predictions of the proportion
of households that would
connect to a new natural gas
distribution system, a second
set of regressions was estimated. The results of these
predictive regressions are presented in Table 9. The sample
* - statistically significant at the 5 percent level; ** - statistically significant at the 1 percent level
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania 13
Table 10. Predicted proportion of households who would connect
size is larger than the sample used for the regression in
Table 8 because some respondents did not answer all of
the questions, and therefore had to be omitted from the
regressions shown in Table 8. The parameter estimates
from the predictive regression are very similar in size to
the parameter estimates from the regression in Table 8.
Based on statistical tests that showed that behavior does
not differ among them, the South Eastern region, South
Central region and Cumberland County are lumped
together into one group in the predictive regression, and
the difference for the North Central region is estimated
relative to that baseline. Again, the regression results
show that respondents from the North Central region
are more likely to connect than respondents from the
other three regions.
Using the regression model in Table 9, it possible to
calculate the proportion of households that meet the
selection criteria who would connect to a new natural
gas distribution system. These proportions can then
be combined with the proportion of households that
are not now currently connected to natural gas who
meet the selection criteria (from Table 2). The result
is a prediction of the proportion of households that are
not now currently connected to natural gas who would
connect if given the opportunity. These are shown in
Table 10 for each region for selected combinations of
upfront costs and payback periods. As Table 10 shows,
the proportion of households that would connect to
14
a new natural gas distribution system is higher in the
North Central region than in the other three regions, is
lower for higher upfront costs, and is lower for longer
payback periods.
The American Gas Association estimates that a typical household that currently uses electric resistance
heat could save between $300 and $1,200 per year
by switching to natural gas heat. The size of the annual savings would depend on the household’s current
system and on its heating needs. Costs to connect to a
natural gas line and install the equipment necessary to
heat using natural gas can vary widely from $3,000 to
$12,000. A typical household might face connection/
conversion costs of $6,000 and an annual savings of
$1,000 (payback time of 6 years). In the North Central
region, 26-32 percent of such households would connect to natural gas if given the opportunity. The corresponding proportions for the other regions are 17-24
percent in the South Central region, 18-25 percent in
the South Eastern region, and 17-23 percent in Cumberland County.
Pennsylvania households are clearly influenced by
both upfront costs and payback time. At a lower upfront
cost and a shorter payback period, the predicted connection rate increases. Would upfront costs lower than
$3,000 and payback times shorter than 3 years result in
even higher connection rates? It is always dangerous to
extrapolate outside the range of the data used to estiThe Center for Rural Pennsylvania
mate a model. However, using the model in Table 9, it
is possible to calculate the predicted connection rate if
the upfront costs are zero and the payback time is zero.
This would represent the highest possible connection
rate that could be achieved. Doing this gives a predicted
connection rate of 35-40 percent for Cumberland County, 35-41 percent for the South Central region, 37-44
percent for the South Eastern region, and 46-50 percent
for the North Central region. Again, these estimates are
extrapolations. But they do suggest that half or more of
Pennsylvania households would not connect to natural
gas under any cost scenario.
a typical Pennsylvania house that faces upfront costs
of $6,000 and annual savings of $1,000, the model
estimated here predicts a probability of connecting that
ranges from 17-23 percent to 26-32 percent. Residents
of the North Central region have the highest probability of connection while residents of the South Central
region and Cumberland County have the lowest probability of connection.
The analysis of the WTP questions showed that the
probability of connection increases as the upfront
costs decrease, and increases as the payback time (in
years) decreases. Still, half or more of the Pennsylvania
households surveyed would not connect to natural gas
under any cost/savings scenario. Reluctance to connect
Conclusions
The survey revealed several important results that are is related to three attitudes/perceptions: worry about
useful to policymakers, natural gas distribution compa- potential future increases in gas prices, the hassle of
installing a new supply line and heating equipment, and
nies, and utility regulators. It should be noted that the
survey results are strictly valid only for households that an inability to afford upfront costs. Technical difficulties of installing new heating pipes or ducts inside the
are located in the four study regions and that are not
home and a desire to wait until current equipment needs
currently connected to natural gas.
to be repaired or replaced were less important factors.
First, most surveyed households were well informed
Whether a specific new natural gas distribution line
about the relative operating costs of different heating
will
result in enough connections to warrant the cost
systems. Second, very few respondents lived in houses
of
constructing
the line will be a decision that must be
that are incapable of being converted to natural gas
made
by
the
distribution
company on a case-by-case
heat due to the inability to install pipes or ducts. These
results suggest that there exists a large potential pool of basis. The answer will depend on how many houses the
new system could serve and on the proportion of those
households who might be interested in connecting to
houses that will choose to connect. The connection pronatural gas to save money on heating.
portion, in turn, will depend on the cost of connection
The WTP questions asked in this survey allow us to
predict the proportion of households that would connect and the annual savings that those houses can anticipate.
to natural gas if it were made available for different sce- The information generated here can help in making that
assessment.
narios regarding upfront costs and annual savings. For
References
Champ, P.A. and R.C. Bishop. Donation Payment Mechanisms and Contingent Valuation: An Empirical Study of
Hypothetical Bias. Environmental and Resource Economics 19(2001), pp. 383-402.
Champ, P.A., R.C. Bishop, T.C. Brown, and D.W. McCollum. Using Donation Mechanisms to Value Non-Use
Benefits from Public Goods. Journal of Environmental Economics and Management 33(1997), pp. 151-162.
Champ, P.A., R. Moore, and R.C. Bishop. A Comparison of Approaches to Mitigate Hypothetical Bias. Agricultural and Resource Economics Review 38(2009), pp. 166-180.
Curtin, R., E. Singer, and S. Presser. Changes in Telephone Survey Nonresponse Over the Past Quarter Century.
Public Opinion Quarterly 69(2005): pp. 87-98.
Ethier, R.G., G.L. Poe, W.D. Schulze, and J. Clark. A Comparison of Hypothetical Phone and Mail Contingent
Valuation Responses for Green-Pricing Electricity Programs. Land Economics 76(2000), pp. 54-67.
Morrison, M., and T.C. Brown. Testing the Effectiveness of Certainty Scales, Cheap Talk, and Dissonance-Minimization in Reducing Hypothetical Bias in Contingent Valuation Studies. Environmental and Resource Economics 44(2009), pp. 307-326.
Poe, G.L., J.E. Clark, D. Rondeau, and W.D. Schulze. Provision point mechanisms and field validity tests of contingent valuation. Environmental and Resource Economics 23(2002), pp. 105-131.
Vossler, C.A., R.G. Ethier, G.L. Poe and M.P. Welsh. Payment Certainty in Discrete Choice Contingent Valuation
Responses: Results From a Field Validity Test. Southern Economic Journal 69(2003), pp. 886-902.
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
15
Appendix 1: Demographic and Socioeconomic Profile
of Pennsylvania Communities
Introduction
This analysis compares demographic and socioeconomic characteristics of Pennsylvania areas with and
without natural gas home heating service. The analysis
was completed to create a baseline measure to assist
with future determinations of the costs and benefits
of expanding Pennsylvania’s natural gas distribution
infrastructure.
Methods
Data Sources
Data used in the analysis are from the 2010 Census
and the 2007-11 American Community Survey (ACS).
Both sources are collected and published by the U.S.
Census Bureau and are available on its website.
Data on the number of school buildings by “block
group” are from the National Center for Education
Statistics, 2010-11 Common Core Data.
Data on the number of hospitals and other health
care facilities are from the Pennsylvania Department of
Health’s Bureau of Health Statistics and Research. The
data were reported for fiscal year 2010-2011.
A municipality may include multiple BGs or parts of
different BGs. This difference is relevant because utilities like gas and cable are not always evenly distributed
throughout a municipality. Residential gas pipelines, for
example, may only extend to a proportion of a municipality, allowing only some households in the municipality the ability to use natural gas. By using BG level
data, the Center was better able to identify smaller areas
with and without gas service.
Applying the Data
Using data from Table B25040, the Center divided
Pennsylvania BGs into two groups: (1) BGs where no
homes were heated with natural gas; and (2) BGs where
one or more homes were heated with natural gas. For
ease of reading, the first group will be referred to as
“non-gas BGs” and the second group as “gas BGs.”
The Center used these two categories to: separate out
those households that do not use natural gas for home
heating thereby reducing the possibility that natural
gas service is available in the area but households are
choosing not to use it; and analyze and compare the
demographic and socioeconomic characteristics of BGs
where homes are and are not using natural gas for heating.
Identifying Areas without Natural Gas
To compare demographic and socioeconomic characteristics of households that use and do not use natural
Data Limitations
gas for home heating, the Center first identified areas
• Access vs. Use: The data used for this analysis only
with and without natural gas service. Since the Center
measure natural gas use, not access. Using smaller
is unaware of any statewide map or database that identilevels of geography (block groups) where no homes
fies areas where natural gas service is and is not availuse natural gas for heating increased the probabilable, it developed its own estimate.
ity that natural gas is not available. Unfortunately,
Its estimate used the 2007-11 ACS data, which are
this probability cannot be quantified since there is
available statewide and for smaller geographic areas.
no known statewide database of the areas with and
Specifically, it used data from Table B25040 (House
without natural gas service. Therefore the analysis
Heating Fuel for Occupied Housing Units) to identify
here represents a reasonable effort to compare areas
areas in which homes are or are not heated by natural
with and without natural gas service.
gas.
• Margin of Error: 2007-11 ACS data were collected
through a random sample of households, so the
Geographic Level of Analysis
data are subject to sampling error. Recognizing this
The smallest geographic areas in which ACS data
limitation, the Center reported data in aggregate
are available are Census Block Groups (BGs). In 2010,
rather than highlighting specific BGs. The data are
Pennsylvania had 9,740 BGs. Their median population
reported as ratios, percentages, means, and medians
was 1,160 and their median geographic size was 0.5
rather than absolute numbers. The exception is the
square miles.
2010 Census data, which are a 100 percent count of
The Center used BGs because they are more numerall households.
ous than municipalities and generally have smaller
populations/households. It is important to note that
BGs do not always conform to municipal boundaries.
16
The Center for Rural Pennsylvania
• No Time Comparisons: Every 10 years the U.S.
Census Bureau re-examines, and in some cases redraws, the boundaries of block groups. Therefore it
is not possible to compare the rate of change.
• Missing Data: Thirty-seven of Pennsylvania’s
9,740 BGs had incomplete or missing data. These
BGs are a combination of large areas of water or
industrial areas that contain no households. Because of their unique characteristics and the lack of
households, these 37 BGs were removed from the
database.
Findings
Block Groups
In 2007-11, 547 (6 percent) of Pennsylvania BGs had
no homes heated by natural gas. As Map 1 shows, 49 of
Pennsylvania’s 67 counties had one or more BGs where
no homes were heated with natural gas. The largest
concentrations (40 percent or more) of these BGs were
in Perry, Schuylkill, and Carbon counties.
In 18 of Pennsylvania’s 67 counties, all BGs have
homes heated by natural gas. These counties include
some of Pennsylvania’s more urban counties, like
Philadelphia and Erie, and some of its more rural counties, like Forest and Potter.
Among non-gas BGs, 68 percent were rural and 32
percent were urban. Among gas BGs, 18 percent were
rural and 82 percent were urban.
Institutions within Block Groups
Among Pennsylvania’s 260 hospitals and health
care facilities, five were located within non-gas BGs,
or 2 percent of the total. The remaining facilities were
located in gas BGs.
Among Pennsylvania’s 2,982 public school buildings,
127 buildings (4 percent) were located within non-gas
BGs and the remaining buildings were located within
gas BGs.
Land Area, Population, and Households
Non-gas BGs encompassed 7,121 square miles, or
16 percent of Pennsylvania’s total land area of 44,743
square miles. The average non-gas BG was 13.2 square
miles.
Gas BGs comprised 84 percent of Pennsylvania’s
land area, or 37,489 square miles. The average gas BG
Map 1: Non-Gas and Gas Block Groups, 2007-11
Note: Non-gas block groups have no homes heated with natural gas and gas block groups have one or more homes heated with
natural gas. The map does not include the 37 block groups without occupied homes. Data source: Table B25040 (House Heating Fuel for Occupied Housing Units) 2007-11 American Community Survey, U.S. Census Bureau.
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
17
was 4.1 square miles. There was a statistically significant difference in the average size of non-gas and gas
BGs.
There were 260,148 households in non-gas BGs, or
5 percent of Pennsylvania’s 5.0 million households in
2010. On average, non-gas BGs had 476 households.
In 2010, gas BGs had 4.7 million households, or 95
percent of the statewide total. Gas BGs had an average
of 520 households.
Approximately 687,000, or 5 percent of Pennsylvania’s 12.7 million residents, lived in a non-gas BG in
2010. The average non-gas BG had a population of
1,256. Gas BGs had 12.0 million residents, or 95 percent of Pennsylvania’s population. The average gas BG
had a population of 1,308. There was no statistically
significant difference in the average population between
non-gas and gas BGs.
Non-gas BGs had a population density of 95 persons
per square mile, and gas BGs had a population density
of 319 persons per square mile.
Poverty and Income
The 2007-11, the poverty rate in non-gas BGs was 9
percent. In gas BGs, the poverty rate was 13 percent.
The average household income in non-gas BGs was
$67,405 and the average in gas BGs was $69,383. The
$1,978 gap between the two BGs was not statistically
significant.
Housing
According to the 2010 Census, there were 308,619
housing units in non-gas BGs, or 6 percent of Pennsylvania’s 5.6 million housing units. The average non-gas
BG had 564 housing units.
In gas BGs, there were 5.2 million housing units, or
94 percent of the statewide total. The average gas BG
had 574 housing units.
Non-gas BGs had a higher housing vacancy rate (16
percent) than gas BGs (10 percent). Most of the vacant
homes in non-gas BGs (65 percent) were for seasonal,
recreational, or occasional use, such as hunting and
fishing camps. In gas BGs, 44 percent of vacant houses
Demographic Characteristics
were either for sale or rent.
In 2010, non-gas and gas BGs had nearly identical
Homeownership was higher in non-gas BGs than gas
percentages of residents under 18 years old (22 perBGs (83 percent and 69 percent, respectively). In noncent), and persons 65 years old and older (15 percent).
gas BGs, 39 percent of householders owned their home
In non-gas BGs, the majority of residents (94 percent) free and clear and 61 percent had a mortgage or loan. In
was white/non-Hispanic. Only 6 percent of the popula- gas BGs, 35 percent owned their home and 65 percent
tion were minorities (non-white and/or Hispanic.) In
had a mortgage or loan.
gas BGs, 21 percent of the population were minorities
In non-gas BGs, the majority of occupied housing
and 79 percent were white/non-Hispanic.
units (79 percent) were single family units; 6 percent
Sixty percent of households in non-gas BGs were
were duplexes or townhouses, 9 percent were mobile
comprised of married couples and 40 percent were
homes and 6 percent were in units in multi-unit buildother types of households, such as non-married couples, ings. In gas BGS, 57 percent of occupied housing units
single parents, or people living alone. Forty-seven
were single family units, 19 percent were townhouses,
percent of households in gas BGS were married couple 4 percent were mobile homes, and 20 percent were
households and 53 percent were other types of houseunits in multi-unit buildings.
holds.
Homes in non-gas BGs were relatively newer than
In non-gas and gas BGs, the same percentage of
homes in gas BGs. For example, in non-gas BGs, 37
households (27 percent) had children.
percent of homes were built before 1960 compared to
Non-gas BGs had a lower percentage of single person 51 percent in gas BGs. Since 2000, however, both nonhouseholds (22 percent) than gas BGs (29 percent).
gas and gas BGs had similar rates of construction (8
On average, non-gas BGs had more persons per
percent).
household (2.56) than gas BGs (2.44).
Forty-eight percent of households in non-gas BGs
Seven percent of the households in non-gas BGs did
used fuel oil and 27 percent used electricity to heat
not speak English at home compared to 12 percent of
their homes. The remaining 25 percent used other fuels.
households in gas BGs.
In gas-BGs, 54 percent of households used natural gas
In non-gas BGs, 19 percent of adults (25 years old
for heating, 19 percent used fuel oil, 19 percent used
and older) had a bachelor’s or graduate degree. In gas
electricity, and 8 percent used other fuels.
BGs, 27 percent of adults had a bachelor’s degree or
higher.
18
The Center for Rural Pennsylvania
Housing Values
In 2007-11, the average value of an owner-occupied
home in a non-gas BG was $208,509. The average
value of a home in a gas BG was $209,065. The $556
difference between the two was not significantly significant.
There was also no statistically significant difference
between non-gas and gas BGs in housing affordability
among homeowners with a mortgage. Affordability was
measured by dividing monthly housing expenses (mortgage payments, insurance, utilities, etc.) by household
income. According to the U.S. Department of Housing
and Urban Development (HUD), households that pay
more than 30 percent of their income for housing are
living in a home that is considered unaffordable. In
non-gas BGs, 65 percent of homeowners with a mortgage are paying less than 30 percent of their income for
housing and 35 percent are paying more than 30 percent. In gas BGs, 67 percent of homeowners are paying
less than 30 percent of their income for housing and
33 percent are paying more. There was no statistically
significant difference between non-gas and gas BGs.
Summary
Some of the some key differences between non-gas
and gas BGs include the following:
• Rurality – non-gas BGs are more likely to be rural
and have lower population densities than gas BGs.
• Race – non-gas BGs are not as racially and ethnically diverse as gas BGs.
• Household types – non-gas BGs have a higher
percentage of married couples, fewer single person
households, and more people living in households
than gas BGs.
• Poverty – the poverty rate in non-gas BGs is lower
than in gas BGs.
• Homeownership – homeownership rates in nongas BGs are significantly higher than in gas BGs.
Similarly, non-gas BGs have more single family
homes than gas BGs.
• Age of housing units – non-gas BGs have a higher
percentage of relatively newer homes than gasBGs.
The similarities between non-gas and gas BGs are
household incomes and housing values among owneroccupied homes.
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
19
TABLE 1: Profile of Non-Gas and Gas Block Groups
20
The Center for Rural Pennsylvania
Table 1: Profile of Non-Gas and Gas Block Groups (continued)
(continued on next page)
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
21
Table 1: Profile of Non-Gas and Gas Block Groups (continued)
22
The Center for Rural Pennsylvania
Table 2: Non-Gas and Gas Block Groups by County, 2007-11
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
23
Table 3: Number of Households in Non-Gas and Gas Block Groups by County, 2007-11
24
The Center for Rural Pennsylvania
Appendix 2: Business and Industry Profile by County
This analysis compares economic and socio-demographic characteristics of Pennsylvania counties according to specific percentages of homes using natural
gas for heating.
The analysis was completed to create a baseline measure to assist with future determinations of expanding
Pennsylvania’s natural gas distribution infrastructure
based on current commercial and industrial activities.
To complete the analysis, the Center for Rural
Pennsylvania first looked at the number of homes that
use natural gas for heating in the state’s 67 counties.
The Center then divided the counties into three groups
based on the percentage of homes heated with natural
gas as follows: counties in which less than 33 percent
of homes are heated with natural gas; counties in which
33 percent to 65 percent of homes are heated with
natural gas; and counties in which 66 percent or more
of homes are heated with natural gas.
The Center then aggregated county level economic
data into these three groups. This aggregation provides
an overview and comparison of employment patterns,
businesses establishments, and market conditions in
counties by the different percentages of natural gas use.
Data Sources
To identify different levels of gas use, the Center
used data from the 2011 American Community Survey, 5-year average (2007-11), specifically data table
B25040. Economic data came from the Pennsylvania
Department of Labor and Industry, Team Pennsylvania, 2011 County Business Patterns, U.S. Census. Data
on socio-demographic conditions came from the U.S.
Census Bureau.
Additional data came from the Pennsylvania Departments of Health, Education, and Corrections.
Table 1: Business and Industry Profile
(continued on next page)
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
25
Table 1: Business and Industry Profile (continued)
26
The Center for Rural Pennsylvania
Table 1: Business and Industry Profile (continued)
Data sources: 1.) 2011 American Community Survey, Five Year Average (2007-11) U.S. Census Bureau. 2.) 2010 Census, U.S. Census Bureau. 3.) PASiteSearch, Team Pennsylvania Foundation and Pennsylvania Department of Community and Economic Development. http://www.pasitesearch.com/. 4.) 2012
Hospital Report and 2012 Nursing Home Report, Health Statistics and Research, Pennsylvania Department of Health. http://www.portal.state.pa.us/portal/
server.pt?open=514&objID=590080&mode=2. 5.) Pennsylvania Department of Corrections http://www.cor.state.pa.us/portal/server.pt/community/institutions/5270 and Federal Bureau of Prisons. http://www.bop.gov/locations/maps/NER.jsp. 6.) Included in the building count are school districts, state juvenile
correctional institutes, intermediate units, comprehensive and occupational career and technology centers and charter schools. Excluded are cyber charter
schools and private schools. Data source: 2012-13 Public School Enrollment Reports. Pennsylvania Department of Education. http://www.portal.state.pa.us/
portal/server.pt/community/enrollment/7407/public_school_enrollment_reports/620541. 7.) 2012 Integrated Postsecondary Education Data System (IPEDS)
National Center for Education Statistics. http://nces.ed.gov/ipeds/. 8.) 2011 County Business Patterns, U.S. Census Bureau. http://www.census.gov/econ/
cbp/. 9.) 2010 average weekly wage data is adjusted for inflation using the CPI-U = 100. Data source: 2012 Annual Quarterly Census of Employment and
Wages (ES202), Bureau of Workforce Information and Analysis, Pennsylvania Department of Labor and Industry. https://paworkstats.geosolinc.com/analyzer/searchAnalyzer.asp?cat=HST_EMP_WAGE_IND&session=IND202&subsession=99&time=&geo=&currsubsessavail=&incsource=&blnStart=True.
10.) 2012 Annual Labor Force, Employment, and Unemployment, Bureau of Workforce Information and Analysis, Pennsylvania Department of Labor and
Industry. https://paworkstats.geosolinc.com/analyzer/searchAnalyzer.asp?cat=HST_EMP_WAGE_LAB_FORCE&session=LABFORCE&subsession=99&ti
me=&geo=&currsubsessavail=&incsource=&blnStart=True.
Map 1: Percentage of Homes Heated
with Natural Gas by County
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
27
Appendix 3: Senate Resolution 29
28
The Center for Rural Pennsylvania
Analysis of Potential Demand for the Extension and Expansion of Natural Gas Distribution Infrastructure in Pennsylvania
29
30
The Center for Rural Pennsylvania
The Center for Rural Pennsylvania
Board of Directors
Chairman
Senator Gene Yaw
Vice Chairman
Senator John N. Wozniak
Treasurer
Representative Garth D. Everett
Secretary
Dr. Nancy Falvo
Clarion University
Representative Rick Mirabito
Dr. Livingston Alexander
University of Pittsburgh
Dr. Theodore R. Alter
Pennsylvania State University
Stephen M. Brame
Governor’s Representative
Taylor A. Doebler, III
Governor’s Representative
Dr. Stephan J. Goetz
Northeast Regional Center
for Rural Development
Dr. Karen M. Whitney
Clarion University
The Center for Rural Pennsylvania
625 Forster St., Room 902
Harrisburg, PA 17120
Phone: (717) 787-9555
Fax: (717) 772-3587
www.rural.palegislature.us
1P1113-600