Analyzing German consumers’ willingness to pay for green electricity tariff attributes: a discrete choice experiment

Table 4 Generalized multinomial logit model in willingness-to-pay space

Variables	GMNL-WTP-space I Basic model	GMNL-WTP-space II Interaction model
Variables	Coefficient (mean)	Coefficient (mean)
Random parameters
Alternative-specific constant (ASC)^a	21.649***	27.983***
Share of green energy	0.022***	0.027***
Switching bonus	0.004**	-0.002
Price guarantee	0.148***	0.063**
Tariff price	− 1 [fixed]	− 1 [fixed]
Non-random parameters^b
Green energy source: solar	0.211**	0.188**
Green energy source: wind	0.196**	0.178**
Green energy source: RE mix	0.059	0.067
Interaction variables
ASC region: east^c		0.502**
ASC × region: south^c		− 1.139***
ASC × region: west^c		0.671***
ASC × town size ^d		− 0.496***
ASC × EEG levy acceptance^e		0.555***
ASC × Green Party identification^e		− 1.038***
Share of green energy × green Party identification^e		0.010***
ASC × food or fuel^e		− 0.646***
ASC × environment is important when buying groceries^e		1.130**
ASC × never switched before^f		− 0.381***
ASC × wish to outsource switching process^e		1.279***
Standard deviations (SD) of parameter distributions
SD ASC	5.654***	4.999***
SD Share of green energy	0.023***	0.020***
SD Switching bonus	0.010***	0.007***
SD Price guarantee	0.088***	0.116***
Scale heterogeneity
Tau	1.014***	1.137***
Goodness of fit measures
Participants/observations	371/4,452	371/4,452
McFadden pseudo-R²	0.309	0.322
Log-likelihood at convergence	− 2716.756	− 2,670.03
Akaike information criterion	5471.512	5408.06

Source: author’s calculations by means of the STATA-command “gmnl” in STATA 14 using 1000 Halton draws
*p < 0.1; **p < 0.05; ***p < 0.001; randomized WTP coefficients with significant SD are assumed to be normally distributed and correlated; the price coefficient was normalized to be log-normal and constrained to − 1
^aBinary coded variable; reference: status quo alternative “No switch.”
^bEffect coded; reference: “Energy source: biogas”
^cEffect coded; reference: “Region: north”
^dThe variable “town size” was divided into five groups, and ranged from “less than 5000 residents” to “more than 500,000 residents”. For a detailed structuring of the groups see Additional file 1: Appendix S3
^eEffect coded; reference: “Participant does not support the queried statement”
^fEffect coded; reference: “Participant switched the electricity tariff at least once before”

ISSN: 2192-0567