Carbon Tax Pass-Through in Colombia: price ceilings and vertical integration

Wesley Blundell and Juan Vélez-Velásquez

2025-05-01

Motivation

• Transportation = 25% of global GHG emissions
• Carbon taxes aim to shift fuel consumption
• Effectiveness depends on who bears the brunt of the tax
  • Colombia: regulated market, vertical integration
  • Effective tax varies across space and time due to ethanol mix
• We ask: how much of the tax is passed to prices, and who pays?

• Transportation accounts for a large proportion of GHG.
• Environmental agencies increasingly using carbon taxes to discourage.
• Effectiveness depends on tax incidence and interaction with other policies.
• Colombia natural setting to study this.
• There’s substantial variation in the effective carbon tax across stations and time due to ethanol blending policies.- - And the market structure is quite different.

Research Questions

• What is the average carbon tax pass-through in Colombia?
• How do price regulation and vertical integration affect pass-through?
• How do competition regulation and environmental regulation interact?

• We focus for now two questions.
• Measure average pass-through, and
• how that changes depending on market regulation and integration.
• But in the future we want to dig deeper into the interaction between regulatory measures.

Colombia´s Gasoline Market

• One refiner: Ecopetrol (state-owned)
• Wholesalers often vertically integrated with retailers
• Exclusive contracts, limited upstream competition
• Central planned economy up until retailer.
• Price regulation
  • Liberalized cities vs. regulated ones
  • “Reference price” published monthly
  • Retail margins capped in regulated cities

• single refiner Ecopetrol dominates refining
• wholesalers are sometimes integrated with stations.
• heavy handed reg supply chain is similar to a central planned soviet economy.
• The Ministry of Mines and Energy divided cities into those it deemed competitive enough and not competitive enough.
• For cities where stations can set prices freely the Ministry publishes a reference price.
• Even in liberalized cities, this anchors expectations. In regulated ones, there’s a binding cap on markups.

Carbon Tax Design

• Introduced in 2017
  • 4 USD per ton of CO₂
  • 0.05 USD per gallon of pure gasoline
• Tax applies only to the fossil portion of gasoline
• Ethanol content (4–10%) reduces effective rate
• Updated with CPI every February

• Tax indtroduced in 2017 as 4 USD per CO2 resulting from combusting fossil fuels.
• A gallon of pure gasoline is taxed proportionally
• Ethanol isn’t taxed.
• At the pumpthe effective tax would be only levied on the part of the mix that is gasoline.
• So if ethanol is 10% of the gallon, only 90% is taxed.
• That creates useful variation in effective tax rates across stations.

Supply chain

• a bit dated but useful
• so you have the one refiner
• couple of ethanol producers.
• Both fuels are mixed at wholesale distribution plants
• trucks take them to individual stations.
• All the costs and taxes before that stage determine the marginal cost of the station.
• Then the station decides a markup over that marginal cost.
• That markup is capped in some cities.

Theory

• Spengler (1950), Weyl and Fabinger (2013), Hong and Li (2017)

  • A wholesaler: \(w=m^w(\mu)(\tilde{c}+\theta^w)\)
    • \(\tilde{c}=c+\tau\)
  • A retailer: \(p=m^r(\epsilon)(w+\theta^r)\)
  • Pass-through rate: \(\rho = \frac{\partial p}{\partial \tilde{c}}\frac{\tilde{c}}{p}\)

• To get a sense on what to expect in terms of the effects of vertical integration and price ceilings on tax pass-through
• we wrote a model based on the seminal paper about vertical integration that was rediscovered by W&F and H& L.
• we define the pass-through rate as

Theory (II)

• Pass-through under VI

\[\rho^V=\bigg( \frac{1}{1+\frac{\partial\epsilon^V}{\partial p^V}\frac{p^V}{\epsilon^V}\frac{1}{\epsilon^V-1}}\bigg) \frac{\tilde{c}}{\tilde{c}+\theta^w + \theta^r} \] - Pass-through under arm lenght’s arrangement:

\[ \rho^A =\frac{1}{1+\frac{\partial \epsilon}{\partial p}\frac{p}{\epsilon}\frac{1}{\epsilon-1}}\frac{1}{1+\frac{\partial \mu}{w} \frac{w}{\mu}\frac{1}{\mu -1}} \frac{\tilde{c}}{\tilde{c}+\theta^w+\frac{\mu -1}{\mu}\theta^r}\]

• We can obtain pass-through rates under arm length’s pricing and vertically integrated pricing
• We have these two terms that do not show up in the integrated case:
• wholesaler markup adjustment (reduce passthrough)
• tax is a largest share of retail cost

What to expect?

• Relative to the vertically integrated case:
  • Arm’s-length → wholesaler shoulders some of the tax
  • Arm’s-length → tax is a larger share of retail cost
  • Pass-through ambiguous in theory

• Theoretically, pass-through could go either way.
• Integration internalize the wholesale margin,
• Tax represents a larger share of cost

Data

• The universe of gas stations
  • Location
  • Characteristics (brand, number of pumps, etc.)
  • (Allegedly) all the prices
• Gasoline transactions between wholesaler and retailer
  • Distance from rack
  • Ethanol mix

• So the upside of all his heavy-handed regulation is that we have great data.
• We observe all stations and their characteristics
• all the prices
• using transaction data we can see which stations are integrated
• and the ethanol mix

Identification Strategy

• Two key sources of effective tax variation:
  • Spatial and temporal changes in ethanol blend
    • Weather shocks
    • International price of sugar
  • Annual February tax adjustment (indexed to inflation)
    • DANE (Colombian BLS) publishes CPI in January

• We get clean estimates on tax PT
• Effective tax varies exogenously because it is function of ethanol mix
• There is also some variation over time. As a result we see in the data mixes that go from 0% to 10%.

Empirical Strategy

\[\begin{multline} \scriptsize ln(Price_{i,t}) =\beta_{0}+\beta_{1}ln(Tax_{i,t})+\beta_{2}Regulation_{i}\cdot ln(Tax_{i,t})+\beta_{3}VI_{i}\cdot ln(Tax_{i,t})\\ \scriptsize +\beta_{4}Regulation_{i}\cdot VI_{i}\cdot ln(Tax_{i,t})+\gamma_{i,Month(t)}+\rho_{i,Year(t)}+\delta_{t}+\varepsilon_{i,t} \end{multline}\]

• Outcome: Log retail price
• Key variables:
  • Regulation dummy
  • Log tax
  • Vertical integration dummy

• To get estimates on PT and how it is modulated by vertical structure and reegulation
• regres log price on log tax with dummies for station’s characteristics
• are they integrated
• do they operate in a market where price ceilings are imposed?
• We also include a suite of fixed effects

Main Result: Overshifting

• Main result: overshifting in pesos
• weird language of PT literature
• Average PT: 0.122, 1 COP additional tax increases prices by 8 COP
• We get negative PT for VI stations in unregulated markets! (1COP tax -0.97COP price)

Emissions Implications

To demonstrate the relevance of these results for determining the emissions impact of a 10% increase in the carbon tax, Figure 2 projects the avoided annual CO2 emissions according to whether stations are subject to price regulation or vertically integrated. Using the 2019 sample year, with a total of 1.66 billion gallons sold, we see that at various gas price elasticities, emissions abatement is highest for stations not integrated with wholesalers and not subject to price regulation, while the lowest level of emissions abatement is found among stations subject to both factors.15 Further, the gap in emissions abatement is larger the more elastic consumers are in terms of gasoline consumption.

Role of Competition

Competition matters. Stations with many nearby competitors are much less able to overshift. Market power plays a key role.

Final remarks

• Carbon tax is overshifted
• Vertical structure matters for incidence
• More work understanding regulation and the role of competition

Our findings show the importance of market structure. Policymakers should consider vertical arrangements and regulation when assessing the burden of carbon pricing.

Thank You

Questions?

Happy to take questions!