Thread by @SylvainCF

Amazing lineup today @TSEinfo conference on Environmental Regulation and Firm Performance.

I’m gonna a try and live twitt the conference. See below for thread.

First talk by U. Wagner on the co-benefits of climate policies.

Climate change policies curbing GES emissions will also, if successful, curb the damages from other pollutants emitted at the same time. These savings might even equal total abatement costs. Wagner and de Preux are the first to look at the ex post cobenefits of EU ETS.

Pb: trading GES permits might generate implicit trades in co pollutants, and possibly an increase in co pollutants:
1/ is it the case?
2/ how has EU ETS altered the spatial distribution of pollution?

Data:

Impressive data work matching pollutants data with accounting data, ETS data, etc, recovering 50% of EU ETS firms and 92% of emissions.

They use two approaches to compute co pollutants:
1/accounting approach using allocated permits as counterfactual.
2/ econometric approach

Accounting approach based on the idea that allocated permits in EU ETS were close to historic emissions. So it is sort of a before after comparison. Most co pollutants decrease but for SOx

Econometric approach: DID interacting Carbon price with share of emissions to be regulated by the ETS as a fraction of historic emission or of sales.

Same result: large increases in SOx due to EU ETS, decreases in other pollutants.

Fascinating results but I’m still struggling to understand where SOx pollution increases and why.

Giulia Pavan now presents a paper estimating the impact of the EU ETS on the productivity (TFP) of Italian firms.

Main results: EU ETS seems to increase TFP.

This seems to plead in favor of the famous Porter hypothesis that environmental regulation improves productivity.

Main comment: could the apparent productivity effects simply be due to differential rates of pass through of the policy cost to consumers by regulated and unregulated firms?

Others have found a positive effect of EU ETS on regulated firms’ profits. Deeply troubling results!

For example Bushnell, Chong and Mansur find that stock prices of regulated firms in the EU ETS DECREASE when permit prices DECREASE, which is unexpected to stay the least.

aeaweb.org/articles?id=10…

Natalia Fabra now presents a paper on the impact of renewables on market power.

Main result: intermittency interacts with market power to make gains from renewables lower than if they were pricing at marginal cost. This is because realized capacity is unknown.

But: compared with a technology with certain capacity (e.g. nuclear), intermittency generally decreases coordination and thus market power.

Raphael Calel is now presenting an estimation of the impact of the EU ETS on firm profitability.

Main result: profits seems to have increased in regulated vs similar unregulated firms.

Main comment (I’m discussing the paper): differential pass through rates could bias results

Meredith Fowlie is presenting work on leakage effects.

A lot of GES emissions are priced today around the world (25%), but a lot of emissions are not.

Consequence: decreases in regulated emissions might be undone by emissions in unregulated emissions. This is leakage.

Economists’ response would be border adjustment tax, but what regulators do in practice is to give away free permits, subsidizing industries at risk of leakage.

Punchline: current policies correctly include emissions intensity and trade penetration, but ignore elasticity of foreign output.

Meredith and Mar use changes in energy prices inside and outside the US and across time and industries to identify these elasticities.

Elasticities seem to be coherent with energy intensity.

Then you can compare optimal subsidies to actual subsidies.

This is what optimal subsidies would look like.

Discussion by Celine Nauges: pb is static model and perfect competition.

My comment: why not use modern empirical trade approaches with differentiated firms?

Mirabelle Muuls is presenting a paper on the impact of the EU ETS on French firms.

Critics of EU ETS argue opposite things:
1/ No impact on emissions
2/ Detrimental impact on exporting firms

Approach is to use matching DID on firms and plant data.

Results:
1/ Decrease in emissions
2/ No change in output
3/ Production seems to reallocate within firms across plants, but not across firms
4/ Probably no leakage

Stefan Lamp is now presenting a paper on the impact of taxing electricity on the performance of German firms.

The tax (EEG) was used to finance renewable energy.

Method: use major reform of the tax that doubled the number of exempts firms in 2012.

Results: increase in electricity consumption after decrease in tax. Some substitution seems to come from gas, but very imprecise. Increase in total emissions.

My main comment: why not use regression kink design: eml.berkeley.edu//~saez/course/…

Now Antoine Dechezleprêtre is presenting a paper on the cost of air pollution in Europe. He especially focuses on the aggregate impact of air pollution on productivity using thermal inversions and wind directions to generate exogenous shifts in air pollution.

Results: huge negative impacts of pollution on production

Nice use of temperature inversions as instruments for pollution.

And of wind direction.

Problem with the instruments: they move concentrations in PM2.5, but also a lot of other pollutants.

Results by sector:

Costs of abatement are much lower than the estimated impact of pollution on GDP.

I’m gonna miss the last presentation because I have to pick my kids from school :)

Live Twitt will resume tomorrow with Michael Greenstone’s keynote lecture.

I’m resuming the live twitt with Michael Greenstone’s keynote.

Michael is the pope or Jedi master of empirical environmental economics.

He’s gonna talk about the central role of information and incentives in determining efficiency of environmental regulation.

Example: Canada dropped out of Kyoto protocol: emissions increased instead of decreasing.

Non compliance with environmental regulation is widespread.

Most programs abating carbon emissions are too costly

Roadmap of the talk is promising, especially by the end where Michael is going to paint the roads for future research. He is generally is extremely insightful in that exercise.

Revealed valuation of reduction in air pollution are large: Clean Air Act Amendments (CAAA) decrease in pm2.5 valued 123billion USD. (From famous Chay and Greenstone paper)

In China, more pm10 decreases life expectancy.

Some wisdom from Michael:
“Data is the worst possible boyfriend or girlfriend that you’ve dated in high school. It is going to disappoint you everyday, you just do not know exactly how."

Using estimates from China, Michael estimates that life expectancy would increase by almost 4 years in India and China if both countries would comply with WHO regulations.

BUT regulations also have substantial costs.

Michael estimates that 15 years of CAAA triggered loss of 600000 jobs and hundreds of billions of dollars in output loss.

Annual economic cost of this regulation on industries is estimated to be around 24 billion dollars.

In China, pollution fees have a substantial negative impact on output.

Now we’re coming to the meat part of the talk: information and incentives. The Laffont Tirole idea that regulated firms have superior information and want to mischaracterize their costs to the regulator is essential here.

How to design robust policies, especially yin LdC?

In practice, regulation is hard and costly, especially the cost of accessing information.

Case in point: corrupt auditors in India reported everyone was in compliance with regulation whereas true inspections detected massive violations.

Just changing the incentives of private auditors (controls + independent funding + random assignation) increase compliance and decreased pollution.

Regulators fine polluters only when they deviate largely from the regulatory threshold: only violations more than 10 times above the standard trigger substantive action by the regulator.

In another experiment, Michael along with others tried to increase the rate of inspections for every firm and to decrease discretion in which firms to inspect. This last part was to curb potential corruption in allocation of inspections.

There was no decline in pollution.

The regulators were actually not corrupt, but using private information to target regulatory effort on large polluters that matter a lot for pollution readings.

At least, that’s how Michael and his coauthors interpret their data using a structural model.

New projects:
1/ Using machine learning to target inspections better
2/ Using better monitoring technology: continuous monitoring
3/ Making pollution emissions public

Reed Walker is bow talking about the impact of corrective taxes on imperfectly competitive industries.

Do industries pass through their cost increases to consumers?

Here is the formula for taxing an input.

Reed is going to estimate all the parameters in this formula for energy inputs, in order to estimate the potential incidence of a carbon tax.

Markups are identified by comparing increases in revenue to increases in inputs: if disproportionate knowing the production function, the wedge is interpreted as markup.

Approach dates back to Hall 1988 and has been extended by Jan de Loecker recently.

Reed uses @TimBartik like instruments to instrument for energy prices, interacting shares of energy sources with national changes in prices

Reed finds that increases in energy prices decrease markups, so pass through is not perfect. Pass through rates are estimated to be around 75%, but there is wide variation across industries, with some industries with pass through rates way above 1!

Main result: accounting for imperfect competition decreases the incidence of taxation on consumers. Incidence is in general close to .5 in their estimates.

Mathias Reynaert suggests that it would be great to be able to decompose the total incidence effects between pass through and substitution in order to understand which markets are going to have strong impacts.

My main comment: would @paulgp and @davidjaeger critique of @TimBartik instruments apply there?

Mar Reguant is now presenting a paper on efficiency and distributional impacts of renewable policies.

Energy tax credits and solar tax credits generally favor rich people that pay taxes, own houses and can make the investment, as @BorensteinS has shown in previous work.

For example, residential consumers bear the burden of German renewable policies through higher electricity prices.

Mar builds a model of electricity production and consumption, with residential, commercial and industrial users, and simulates a carbon tax and a feed-in tariff and a renewable portfolio standard imposing a minimum share of electricity produced from renewables.

Key question is whether typical Ramsey pricing (taxing in elastic consumers) can rationalize the observed price structure taxing residential consumers.

Mar assumes the elasticities of consumers consumption is lower than other sectors. This is going to bring her towards Ramsey pricing.

Mar describes the supply curve as a cost minimizing decision.

Results:
1/ Mandates and subsidies leverage more on changes in the production side and tax decreases consumption more.
2/ Leakage seems to be the main justification for decreasing prices for industrials.

Ken Gillingham is now talkin about the impact of fuel economy standards on the structure of the car market, called CAFE standards in the US.

Problem is that CAFE lowers weight of vehicles, which is associated with increased fatalities.

The National Academy of Sciences failed to come to a shared conclusion on this issue.

The main problem is that of weight dispersion: big cars are more dangerous when facing lighter cars.

Final effect of CAFE on fatalities is thus complex to assess.

Weight decreased after CAFE was ramped up, but not uniformly

Ken uses Firpo’s Regression influence function to estimate the impact of regulation on unconditional quantiles of the weight distribution.

Pb: how to estimate stringency of regulation?

Here: stringency is firm specific deviation from pre Cafe standard.

Really cool historical data on cars markets in the US since 1948 to 2011.

Results: this is mainly the weight of smaller cars that react to the stringency of the standard.

This increases dispersion.

Driving a lighter vehicle also does increase fatalities.

But they also find that average lighter vehicles decrease fatalities.

Now you can recompute the fleet in the absence of Cafe and estimate total fatalities. Same number of accidents with different weights.

Results: CAFE does not kill, the average weight effect dominates.

The magnitude of the positive effects is enough to overturn cost benefit analysis of CAFE standards.

Main comment by Isis Durrmeyer: reduced form estimates of weight adjustment to standard ignore price responses which might alter market shares and thus fleet composition.

End of thread.

Not end of conference, but I have some appointments this afternoon (I have to do some actual work at some point ;)

Apologies to the speakers from this afternoon.

Hope you have enjoyed the live-twitt. It was sure fun to do. Great inspiring research!