Energy Investments – A Discrete Choice Experiment with Manufacturing Executives

1. Discrete Choice Experiment (DCE)

The DCE has two components:

The first explores Investment decision preferences: CEOs choose between hypothetical investment options (X vs. Y) that vary along four dimensions:

Initial investment cost (€250k–€1m)
Annual savings (€50k–€200k)
Annual emissions reductions (100–500 tons)
Uncertainty of savings (+/– €10k–40k)

The second component aims to elicit Public funding preferences: CEOs choose between hypothetical funding schemes differing on six attributes:

Type of incentive (tax credits, loans, grants)
Maximum support (25–75% of costs)
Administrative burden (low–high)
Flexibility of use (flexible vs. restricted)
Duration (1–6 years)
Environmental goals (none, sustainability, carbon reduction)

To reduce complexity, only 4–6 attributes are used per choice task. Respondents complete two choice scenarios for investments and two for funding schemes. Random assignment into 10 blocks (stratified by firm size) ensures statistical efficiency.

2025-09-15

2025-09-30

Each participant is provided with four DCE tasks (4 = 2 x 2) two for investment decisions and other two for funding. Specifically, in all tasks each participant is asked to express a preference between two options. These choices are our primary outcome. There are other two primary outcomes: (1) the stated social tipping point on a scale from 0 to 100 as measured by question Q26; and (2) the self-reported willingness to pay for additional information helping to improve “greenness” of their firm’s products on a scale from 0 to 10 depending on the number of consumers who are sensitive to it, as in question Q27.

The DCEs enable to compare decisions across CEOs with different personal and behavioural characteristics or from different types of firms. Specifically, we analyse whether CEOs make significantly different choices depending on the following binary variables: the size of their company (Q5, medium vs large), risk aversion (Q18, below vs above the median), time preferences (Q19, below vs above the median), environmental values (Q20, below vs above the median), climate change personal duty (Q23) and customer green perceptions (Q28, below vs above median).

We use algorithms in the EDT package in Python to split respondents into ten blocks. As from the documentation, EDT is a Python-based tool to construct D-efficient designs for DCEs based on the widely used D-efficiency criterion (see Kuhfeld, 2005) but with speed improvements. This approach requires us to set some priors for the model parameters, and a given number of blocks. We decided to do ten blocks, yielding an expectation of 100 respondents per block. Blocks are assigned stratifying by firm size (50 medium 50 large).

Computer randomisation.

Respondent.

No

1,000 firms (500 medium and 500 large)

We aim to recruit one executive per firm, with a total of 1,000 respondents.

We split respondents in 10 blocks, 100 persons per block.