Chapter 4. Analysis of Consumer Propensity for Bundled Energy Services
C. Estimation of the Utility Function
The observable consumer utility in relation to the provider of energy, including electricity defined in formula (2) can be illustrated as follows:
Formula (6)
Here, ASC (alternative specific constant) is a dummy variable expressed in a certain alternative that represents the case in which a respondent chooses the current alternative instead of virtual alternatives. When a respondent chooses to receive electricity from KEPCO instead of from an imaginary energy provider, ASC is 1; in other cases, it is 0. FGAS, FCOM, and FHEAT are dummy variables that represent energy providers, and their values are determined using KEPCO as a standard. For the purpose of this study, FGAS, FCOM, and FHEAT are not ordinary dummy variables but effect- coded dummy variables. The differences between dummy variables and effect-coded dummy variables are as follows:
With ordinary dummy variables, the estimation value of the variables, which are considered to be reference values, is included in the constant, making it difficult to estimate the exact reference value. In other words, each dummy variable
25 Haab and McConnell (2002).
26 Train (2003).
expresses the difference from the reference value, but it is impossible to determine the exact reference value. To remedy this shortcoming, we use effect-coded dummy variables. In this study, the effect-coded dummy is assigned a value of 1 to represent the presence of the dummy variable in the relevant group; a value of 0 to represent the presence of the dummy variable in an irrelevant group; and a value of -1 to represent the presence of the dummy variable in the reference group.
Once the coefficient value for each dummy variable is estimated, we can estimate the coefficient of the reference group.27 Table 4-4. Comparison of Ordinary Dummy Variables and Effect-coded Dummy Variables
Dummy Variables Effect-coded Dummy Variables Electricity
provider FGAS FCOM FHEAT FGAS FCOM FHEAT
Gas
company 1 0 0 1 0 0
Internet/
Telecommu nications company
0 1 0 0 1 0
District heating company
0 0 1 0 0 1
KEPCO 0 0 0 -1 -1 -1
BGAS, BCOM, and BSOL are dummy variables that express the existence or non-existence of product bundles. First, BGAS is given a value of 1 if the given electricity provider bundles gas and heating products with its electricity service and a value of 0 if no such bundle is offered. BCOM and BSOL are also given a value of 1 if the electricity provider bundles Internet/telecommunication or solar system leasing services with its electricity service and a value of 0 if no such bundles or services are offered. PRICE refers to the price of electricity per kilowatt-hour. Since it is obvious that an increase in the electricity price would lead to a decrease in consumer utility, we expect the coefficient of PRICE to be a negative number. Table 4-5 below shows a summary of each variable in the utility function.
Table 4-6 presents the foundation statistics for each variable. Since the current situation is always included in each multiple-choice question, the ASC has an average of 0.25.
Table 4-5. Definition of Variables in the Utility Function
Variable Definition
ASC The value is 1 if the chosen alternative is the current situation; if not, 0.
FGAS
If the chosen alternative is a gas company, the value is 1;
Internet/telecommunications/district heating company, 0;
and KEPCO, -1 FCOM
If the chosen alternative is an Internet/telecommunications company, the value is 1;
gas/district heating company, 0; and KEPCO, -1 FHEAT
If the chosen alternative is a district heating company, the value is 1; an Internet/telecommunications/gas company, 0; and KEPCO, -1
BGAS If the chosen alternative offers gas/heating service, the value is 1; if not, 0
27 Kugler et al. (2012).
BCOM
If the chosen alternative offers Internet/telecommunications service, the value is 1; if not, 0
BSOL If the chosen alternative offers solar system leasing service, the value is 1; if not, 0
PRICE Price of electricity offered by the alternative (in KRW per kWh)
Table 4-6. Basic Statistics of Variables Variable Average Standard
deviation Min. Max.
ASC 0.25 0.433029 0 1
FGAS -0.12562 0.739434 -1 1
FCOM -0.12562 0.739434 -1 1
FHEAT -0.125 0.739956 -1 1
BGAS 0.374383 0.483982 0 1
BCOM 0.406559 0.49121 0 1
BSOL 0.374383 0.483982 0 1
PRICE 259.3904 25.58963 225 325
The utility function defined in formula (6) was estimated from the rank-ordered logit model, and the results are listed in Table 4-7. The log likelihood of the prediction model is –9943.848, and the chi-squared test proved the model to be significant.
Table 4-7. Estimated Values of the Utility Function Variable Coefficient Standard
error z P>z
ASC -0.3316 0.0934 -3.5500 ***
FGAS -0.5781 0.0653 -8.8500 ***
FCOM -0.4443 0.0684 -6.4900 ***
FHEAT -0.2608 0.0626 -4.1600 ***
BGAS -0.3392 0.0355 -9.5600 ***
BCOM -0.3858 0.0415 -9.2900 ***
BSOL 0.0508 0.0320 1.5900 - PRICE -0.0078 0.0006 -12.4500 ***
Log Likelihood -9943.848
AIC: 1.536
BIC: -102763
Note: ***, **, and * are significant at levels of 1, 5, and 10 percent, respectively.
Regarding the significance of each variable, all variables except for BSOL were found to be statistically significant at the one-percent level, from which we can presume that the type of energy provider and their bundled offers have a significant impact on consumer utility. The coefficient for ASC was -0.3316, which was significant at the one-percent level, showing that the respondents preferred switching to a new electricity provider instead of staying with KEPCO.
From this we can glean that consumers expect improved services, compared to the services provided by KEPCO, with the opening of the retail electricity market.
The coefficients of effect-coded variables are negative and statistically significant at the one-percent level, which means that respondents preferred receiving their electricity from KEPCO. The coefficients for KEPCO can be restored using the estimated coefficients of the effect-coded variables. The largest coefficient value was 0.3208, in the event KEPCO supplies electricity, followed by 0.0599 if the electricity is supplied by district heating companies, -0.1235 if supplied by Internet/telecommunications companies, and -0.2572 if supplied by gas companies. This represents the respondents’ preferences for having a company specialized in electricity supply as their electricity provider.
The coefficients of the variables for bundled offers, BGAS and FCOM, were both negative, meaning that bundled offers failed to provide utility to consumers. In general, respondents did not prefer one company providing electricity bundled with gas/heating or Internet/telecommunications services. This means that, without the discounted price of bundled services, consumers did not find the bundled provision of electricity and other services attractive.
On the other hand, the coefficient for BSOL was not significant at the 10-percent level, which meant that the solar system leasing service did not provide additional utility to consumers, despite its advantage of reducing electricity bills.
Since the solar system leasing service was not considered as a factor capable of influencing electricity prices in this study, it seemed to have no effect on consumer utility, which is in line with the results of the question regarding whether the respondents would switch their electricity providers to ones that provided solar system leasing services, as the number of respondents who answered positively closely matched the number of respondents who answered negatively.
The coefficient of PRICE was negative and statistically significant at the one-percent level, as was expected. Consumer utility decreases as the unit price of electricity increases.
Now, using the estimated utility function, we can estimate the marginal willingness to pay (MWTP) for each attribute.
The MWTP is compensating variation, which is the amount consumers are willing to pay in order to maintain the same level of utility when the degree of an attribute changes. The MWTP for an attribute can be deduced from the following formula:28
Formula (7)
For KEPCO’s provision of electricity, the MWTP was KRW 41.17, but for electricity provided by a gas company, the MWTP was KRW -33.01. Regarding bundled energy services, the MWTP was KRW -43.53 for heating/gas bundled with electricity and KRW -49.51 for Internet/telecommunications bundled with electricity, showing that customers had negative views on the value of bundled services. Negative willingness to pay signifies the amount of compensation it would take for consumers to be willing to choose that option. Therefore, it could be understood as their willingness to accept. Since MWTP was measured as the price of electricity per kilowatt-hour, we calculated the amount consumers were willing to pay for the use of a monthly average of 300 kWh of electricity. Table 4-7 shows the MWTP for each attribute per month.
28 Train (2003).