Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Consequentiality and the Willingness-To-Pay forRenewables: Evidence from Germany

Mark Andor1 Manuel Frondel1,2 Marco Horvath1,2,3

1RWI Essen 2Ruhr University Bochum 3RGS Econ

15th IAEE European Conference, ViennaSeptember 6th, 2017

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 1

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Overview

1 Introduction

2 Data and Experimental Design

3 Descriptive Results

4 Methodology

5 Results and Policy Implications

6 Summary and Conclusion

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 2

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Motivation

Non-market goods (e.g. reductions in pollution) are valued on basis of statedpreferencesContingent Valuation Methods:

1 Single Binary Choice2 Open-Ended Method

Stated preference studies may suffer from hypothetical biasTo reduce this bias:

Ex ante: Consequential ScriptEx post: Question for political consequentiality

We investigate the discrepancy in WTP bids across Single Binary Choice andOpen-Ended valuation formats while simultaneously controlling for politicalconsequentiality

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 3

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Survey

Elicitation of WTP for renewable energy using a large-scale survey (amongmore than 7,000 German households)Renewable energy is financed by a surcharge on the electricity bill (EEG Levy)All survey participants get a brief introduction, indicating:

The share of renewable energy in electricity production in 2015: 28%Germany’s target by 2020: 35%The 2015 EEG Levy: 6.17 cents/kwhInformation on the cost of the EEG Levy for an average household

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 4

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Contingent Valuation Formats

Single Binary ChoiceWould you be willing to pay an additional X cents on the per kilowatt hoursurcharge in order to reach the target of 35% renewable energy in the electricitymix by 2020?(X is randomly replaced with either 1, 2, or 4)

Advantage of Single Binary Choice Format: No incentive to strategicallyover- or understate WTP

Open-Ended FormatIn order to reach the target of 35% renewable energy in the electricity mix inGermany, what would the maximum increase of the per kilowatt hour surcharge incents be that you would be willing to pay?

Advantage of Open-Ended Format: Provides information on the whole rangeof respondents’ WTP

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 5

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Consequential Script

Consequential ScriptWe would like to point out that this survey is part of a research project on behalfof the German Federal Ministry of Education and Research (BMBF). The resultsof this survey will be made available to policy makers and serve as a basis forfuture decisions, especially with respect to the future level of the surcharge for thepromotion of renewable energy technologies (EEG Levy). To reach meaningfulconclusions, it is therefore important that you provide exactly thewillingness-to-pay you would actually would be willing to pay at most.

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 6

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Split-Sample Survey Design

Table: Experimental Design: Shares and Number of Observations in TreatmentGroups

Consequential ScriptNo Yes Total Shares

Single Binary Choice

1 Cent 552 534 1,086 33.8%2 Cents 525 537 1,062 33.1%4 Cents 528 536 1,064 33.1%Total 1,605 1,607 3,212 52.7%

Open-Ended 1,401 1,479 2,880 47.3%Total 3,006 3,086 6,092 100.0%Shares 49.3% 50.7% 100.0% –

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 7

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Political Consequentiality

Question about perceived political consequentialityHow likely do you believe that results of surveys like the present one influencepolicy decisions on the amount of the surcharge for the promotion of renewableenergy technologies (EEG Levy)?

Respondents who answer “Very unlikely” are allocated to the inconsequentialgroup (about 40% of all respondents) the rest is allocated to theconsequential group (following Vossler and Watson, 2013)Economic theory suggests consequentiality is needed for incentivecompatibility (Carson and Groves, 2007; Vossler et al., 2012)

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 8

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Descriptives

Means

Open- Single BinaryVariable Variable Definition Ended ChoiceAge Age of respondent 55.2 55.4Female Dummy: 1 if respondent is female 0.352 0.329Children Dummy: 1 if respondent has children 0.704 0.703College Degree Dummy: 1 if household head has a college degree 0.321 0.312Script Dummy: 1 if household received a consequential script 0.500 0.500Consequentiality Dummy: 1 if respondent believes that surveys

influence the political decision making 0.591 0.608Low income Dummy: 1 if net monthly household income

is lower than e1,200 0.073 0.072Medium income Dummy: 1 if net monthly household income

is between e1,200 and e2,700 0.361 0.381High income Dummy: 1 if net monthly household income

is between e2,700 and e4,200 0.293 0.275Very high income Dummy: 1 if net monthly household income

exceeds e4,200 0.148 0.151Missing income Dummy: 1 if respondent did not disclose her income 0.125 0.1211 Person Dummy: 1 if # household members equals 1 0.269 0.2752 Persons Dummy: 1 if # household members equals 2 0.489 0.4723 Persons Dummy: 1 if # household members equals 3 0.132 0.130> 3 Persons Dummy: 1 if # household members >3 0.109 0.123

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 9

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Open-Ended and Single Binary Choice Values

We convert open-ended bids to discrete values for the comparisonOpen-ended responses are randomly allocated to 3 different groups (1, 2, and4 cents)The respective bids are then converted into a binary variable assuming thatrespondents would have accepted a randomly given increase if their WTP bidwere to be at least as large as the respective increase (Balistreri et al, 2001)

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 10

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Descriptive Comparison

Table: Acceptance Rates of a Rise in the Promotion Cost of RenewableTechnologies across Elicitation Formats

Single Binary Choice Open-Ended

Number of Share of Yes Number of Share of YesObservations Responses Observations Responses t-Stat

1 Cent 1,086 53.6% 951 70.5% -7.93***2 Cents 1,062 46.3% 978 57.4% -5.01***4 Cents 1,064 33.7% 951 33.7% 0.03Total 3,212 44.6% 2,880 53.9% -7.26***Note: ∗ denotes significance at the 5 %-level, ∗∗ at the 1 %-level,and ∗∗∗ at the 0.1 %-level, respectively.

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 11

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Regression Model

Yesi = β0 + β1SingleBinaryChoicei + β2 2 Centsi + β3 4 Centsi + β4 Scripti

+β5Consequentialityi + β6(Consequentialityi ∗ SingleBinaryChoicei )+δT xi + εi ,

Yes: Dummy: 1 if individual i accepts a given increase in the EEG LevySingleBinaryChoice: Dummy: 1 if i received the Single Binary Choice question, rather thanthe Open-Ended question2 Cents and 4 Cents: Dummies: 1 if increase was 2 or 4 cents, rather than 1 centScript: Dummy: 1 if i received Consequential ScriptConsequentiality: Dummy: 1 if i believes that surveys influence the political decisionmakingx: Socio-economic characteristics

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 12

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Endogenous Switching Regression Model – First Stage

Switching Regression Model copes with potential endogeneity ofconsequentialityFirst Stage divides respondents into two regimes:

Consequentialityi = 1 if γT · zi ≥ ui ,Consequentialityi = 0 otherwise,

where z includes factors that may affect whether a respondent believes inconsequentiality or notγ of first stage can be estimated by standard probit methods

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 13

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Endogenous Switching Regression Model – First Stage

First Stage identification usually requires an exclusion restrictionWe use two exclusion restrictions:

1 Dummy that indicates whether a respondent took longer than the medianduration to finish the survey

2 Locus of ControlThose believing that life’s outcomes are due to their own efforts have aninternal locus of control, while those believing that outcomes are due toexternal factors have an external locus of control (Gatz and Karel, 1993)Index is constructed following Cobb-Clark and Schurer (2013)

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 14

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Endogenous Switching Regression Model – Second Stage

Depending on consequentiality status, the second-stage equations are given by:

WTP1i = βT1 · x1i − σ1u · IVM1i + ε1i , if Consequentialityi = 1,

WTP0i = βT0 · x0i + σ0u · IVM0i + ε0i , if Consequentialityi = 0,

where IVM are variants of the inverse Mills ratios:

IVM1i := φ(γT · zi )Φ(γT · zi )

, IVM0i := φ(γT · zi )1 − Φ(γT · zi )

For the second stage we use the predicted values IVM1i and IVM0i using theprobit estimates γ of the first stage

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 15

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Linear Probability Model and Probit Estimation Results

Linear ProbitProbability Model Marginal Effects

SingleBinaryChoice -0.190*** (0.019) -0.202*** (0.020)2 Cents -0.103*** (0.016) -0.100*** (0.015)4 Cents -0.263*** (0.015) -0.258*** (0.014)Script -0.005 (0.013) -0.005 (0.013)Consequentiality 0.208*** (0.019) 0.194*** (0.018)Consequentiality * SingleBinaryChoice 0.124*** (0.026) 0.138*** (0.026)Female 0.079*** (0.014) 0.079*** (0.014)Children -0.052** (0.017) -0.053** (0.017)Age 0.002** (0.001) 0.002** (0.001)College Degree 0.063*** (0.014) 0.062*** (0.014)High income 0.008 (0.020) 0.006 (0.020)Medium income -0.024 (0.021) -0.027 (0.021)Low income -0.036 (0.033) -0.039 (0.033)Missing income -0.058* (0.026) -0.059* (0.026)1 Person 0.005 (0.027) 0.004 (0.027)2 Persons -0.045* (0.023) -0.045* (0.023)3 Persons -0.025 (0.025) -0.025 (0.025)Constant 0.461*** (0.038) – –Number of Observations: 5,249 5,249Note: Standard errors are in parentheses, ∗ denotes significance at the 5 %-level, ∗∗ at the 1 %-level,and ∗∗∗ at the 0.1 %-level, respectively.

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 16

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Endogenous Switching Regression Estimation Results

First Stage Second Stage

Consequentiality = 0 Consequentiality = 1SingleBinaryChoice 0.048 (0.036) -0.193*** (0.020) -0.064*** (0.017)2 Cents -0.006 (0.044) -0.087*** (0.026) -0.113*** (0.020)4 Cents -0.084 (0.044) -0.221*** (0.025) -0.273*** (0.022)Script 0.091* (0.036) -0.031 (0.021) -0.016 (0.018)Female 0.118** (0.040) 0.090*** (0.025) 0.046* (0.021)Children -0.083 (0.050) -0.033 (0.028) -0.041 (0.024)Age 0.0001 (0.002) 0.001 (0.001) 0.002** (0.001)College Degree 0.298*** (0.041) 0.001 (0.033) 0.021 (0.029)High income 0.075 (0.057) -0.028 (0.033) 0.005 (0.026)Medium income -0.027 (0.059) -0.049 (0.034) 0.008 (0.028)Low income -0.162 (0.094) 0.007 (0.055) -0.019 (0.049)Missing income -0.219** (0.073) 0.017 (0.046) -0.061 (0.042)1 Person 0.138 (0.077) -0.033 (0.042) -0.013 (0.039)2 Persons 0.058 (0.065) -0.047 (0.035) -0.065* (0.032)3 Persons 0.205** (0.071) -0.039 (0.042) -0.070 (0.038)More time 0.166*** (0.038) – – – –Locus of Control -0.012*** (0.003) – – – –IVM0 – – 0.191 (0.118) – –IVM1 – – – – -0.322* (0.139)Constant 0.248* (0.113) 0.345** (0.114) 0.899*** (0.118)Number of Observations: 5,104 1,999 3,105Note: Standard errors are in parentheses, ∗ denotes significance at the 5 %-level, ∗∗ at the 1 %-level,and ∗∗∗ at the 0.1 %-level, respectively.

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 17

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

WTP for Renewable Energy in Germany

0.2

.4.6

.81

Pol

icy

Sup

port

0 2 4 6 8 10Levy-increase in ct/kwh

Open Ended Single Binary Choice

Willingness-to-pay

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 18

Introduction Data and Experimental Design Descriptive Results Methodology Results and Policy Implications Summary and Conclusion

Summary and Conclusion

We find further evidence on the discrepancy between the outcomes of SingleBinary Choice and Open-Ended valuation methodsContrasting with the literature we find higher WTP values for theOpen-Ended methodWe find a positive relationship between consequentiality and WTPConsequentiality furthermore seems to reduce the discrepancy between SingleBinary Choice and Open-Ended contingent valuation

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 19

Table: Balancing in Explanatory Variables Across Treatment Groups

Open-Ended Single Binary Choice

Format All 1 Cent 2 Cents 4 CentsAge 55.21 55.37 55.39 55.57 55.16Female 0.352 0.329 0.319 0.343 0.326Children 0.704 0.703 0.704 0.714 0.690College Degree 0.321 0.312 0.317 0.307 0.312Script 0.500 0.500 0.499 0.500 0.500Consequentialiy 0.591 0.608 0.620 0.625 0.5791 Cent 0.331 0.333 1 0 02 Cents 0.340 0.333 0 1 04 Cents 0.330 0.333 0 0 1Low income 0.083 0.082 0.080 0.088 0.078Medium income 0.412 0.433 0.437 0.428 0.434High income 0.334 0.313 0.304 0.303 0.331Very high income 0.169 0.172 0.179 0.181 0.1571 Person 0.269 0.275 0.273 0.269 0.2822 Persons 0.489 0.472 0.486 0.478 0.4533 Persons 0.132 0.130 0.123 0.125 0.143>3 Persons 0.109 0.123 0.118 0.128 0.122More time 0.510 0.494 0.503 0.508 0.470# of Observations 3,517 3,524 1,174 1,175 1,175

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 20

0.2

.4.6

.81

Pol

icy

Sup

port

0 2 4 6 8 10Levy-increase in ct/kwh

Open Ended Single Binary Choice

Willingness-to-pay if Respondentsdo not believe that the Survey is consequential

0.2

.4.6

.81

Pol

icy

Sup

port

0 2 4 6 8 10Levy-increase in ct/kwh

Open Ended Single Binary Choice

Willingness-to-pay if Respondentsbelieve that the Survey is consequential

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 21

Table: Acceptance Rates of a Rise in the Promotion Cost of RenewableTechnologies when Elicitation Formats are Crossed with the Consequential Script

Single Binary Choice Open-Ended

ConsequentialScript No Yes No Yes

# of Share of # of Share of t Statis- # of Share of # of Share of t Statis-Obs. Yes Obs. Yes tics Obs. Yes Obs. Yes tics

1 Cent 552 53.8% 534 53.4% 0.14 465 70.1% 487 70.8% -0.252 Cents 525 47.1% 537 45.6% 0.46 479 57.6% 499 57.1% 0.164 Cents 528 34.1% 536 33.4% 0.24 457 31.3% 493 35.9% -1.50Total 1,605 45.1% 1,607 44.1% 0.56 1,401 53.2% 1,479 54.6% -0.75Note: ∗ denotes significance at the 5 %-level, ∗∗ at the 1 %-level, and ∗∗∗ at the 0.1 %-level, respectively.

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 22

Table: Acceptance Rates of a Rise in the Promotion Cost of RenewableTechnologies when Elicitation Formats are Crossed with Consequentiality

Single Binary Choice Open-Ended

Consequentiality No Yes No Yes

# of Share of # of Share of t Statis- # of Share of # of Share of t Statis-Obs. Yes Obs. Yes tics Obs. Yes Obs. Yes tics

1 Cent 406 32.0% 666 66.5% 11.65*** 380 53.2% 561 81.8% 9.91***2 Cents 398 21.6% 651 61.4% 13.61*** 380 42.9% 592 66.6% 7.48***4 Cents 446 13.0% 603 49.3% 13.24*** 391 23.0% 552 41.1% 5.90***Total 1,250 21.9% 1,920 59.4% 22.29*** 1,151 39.5% 1,705 63.3% 12.87***Note: ∗ denotes significance at the 5 %-level, ∗∗ at the 1 %-level, and ∗∗∗ at the 0.1 %-level, respectively.

Marco Horvath Consequentiality and the Willingness-To-Pay for Renewables September 6th, 2017 23