Strategic Responses to Carbon Pricing: Evidence from Firms' Beliefs

We implement a firm survey to elicit firm managers' responses to carbon pricing scenarios. Our survey is designed to elicit production function parameters and biased beliefs. We conduct the survey on a large representative sample of Swiss firms. We vary scenarios within and across firms. Our main identification strategy comes from within-firm variation in carbon pricing scenarios. We use the between-firm variation to assess the overall response to the level of the carbon price.

2025-06-18

2025-11-30

The core module of our survey asks about the effects of carbon pricing on production, price, labor, profit, and energy. We use these outcome variables to assess the effects of carbon pricing. We also elicit firm managers' beliefs about the impact of carbon pricing on other firms' production and pricing decisions. We also use these responses as outcome variables.

Using the responses to the questions about production and price, we plan on estimating a structural model that allows us to back out strategic interaction parameters. These parameters are also outcomes of interest. We plan to investigate their heterogeneity and explain their variation using firms' observables.

We also plan to use firm managers' beliefs about other firms' outcomes to assess biased beliefs. We plan on aggregating beliefs about own-firm outcomes at the sectoral and national level and comparing those averages with beliefs about sectoral or economy-wide impacts. Note that we are planning to use survey weights to aggregate at the sectoral and national levels. Survey weights are needed given that we are planning to oversample large firms.

We also elicit firm managers' support for carbon pricing. We will use those responses in a secondary analysis to assess support for a particular carbon price.

We have structured our survey so we are able to relate firm managers' support for carbon pricing to structural model that map the level of support for carbon pricing to profit. We will estimate a function that models a firm's carbon support as a function of any carbon price.

Our main experimental design consists of presenting carbon pricing scenarios where we vary the scope of the policy. Specifically, we provide firm-specific estimate of CO2 emissions and we present three carbon tax scenarios. Each scenario varies the scope of tax, namely: national, sectoral, or firm-specific. For each scenario, we then ask managers’ beliefs about the expected impact of such a tax on other- and own-firm outcomes. We focus on production and final prices. These different scenarios allow us to estimate parameters that capture how strategic firms are with respect to carbon price, and how biased they are in their beliefs.

We also randomize the tax level across survey respondents to assess whether the level of the tax has the expected impact. We use this variation only for carrying a reduced-form estimation of the average elasticity of different outcomes with respect to a carbon price.

See our detailed pre-analysis plan for further details.

Randomization of low and high tax levels (100 CHF or 300 CHF per ton of CO2) is done across subjects using the online survey instruments.

Survey respondents

No

NA

Our goal is to obtain at least 1000 complete responses. Given that there is uncertainty about response rates to the survey, we will proceed as follows. Survey invitees are drawn from three sources: (i) the KOF survey panel, (ii) respondents from an earlier survey by a subset of the same researchers in 2023; (iii) the Orbis database.\\ Based on small pilot studies, we anticipate the completion rates to differ across these three sources and across different firm sizes. We expect completion rates of at most 2\% for micro-sized firms and 7\% for small, medium and large firms. Based on these assumptions, and to obtain a sample that can be made representative across firm sizes, we will invite in a first round 27,470 firms: 17,164 micro-sized firms (less than 10 employees), 5,257 small firms (10-49) and 5,049 medium-size or large firms (50 or more employees). This will include all the firms from sources (i) and (ii), while the remaining firms will be drawn from Orbis (source (iii)). If, after the first round of invitations, we remain below 1000 responses, we will keep inviting additional firms from Orbis until we reach 1000 complete responses or run out of firms to invite. We have stratified the sample to ensure an appropriate coverage of firm size and sector.

We vary the tax level across subjects. We do a 50-50% randomization as respondents enroll in the survey. Given our target is to have 1000 responses, we will have about 500 respondents in each treatment arm.