The rolling road¶

A car is strapped to a rolling road. It is driven through a fixed sequence of accelerations, steady speeds and stops, the same sequence every time, and the exhaust is measured. What comes out is a figure in grams per kilometre, and the figure is correct. It states what that car emitted over that sequence in that laboratory, and it goes on being correct after everybody has worked out that the sequence resembles nothing anybody drives.

For most of the last two decades that figure was how Europe established compliance with the emission limits. Emissions of light-duty vehicles were measured, for type approval, by laboratory test on a chassis dynamometer running the New European Driving Cycle.

The test is also a specification¶

The Commission’s own research centre described what a laboratory test does, in the first sentence of a report published in 2011. Emissions testing in the laboratory “forms an essential part of the European type-approval procedure”, yields reproducible and comparable data, and provides clear design criteria for vehicles that have to comply with applicable emission limits.

A test that is reproducible enough to regulate against is reproducible enough to build towards, and the two are not separable activities. An engineer optimising a calibration for the cycle is not evading the rule. The cycle is the rule.

Twelve cars, four routes¶

The same report put portable measurement equipment in the boot and drove twelve light-duty vehicles, all compliant with Euro 3 to 5 limits, over rural, urban, uphill and downhill, and motorway routes.

The diesel cars’ route-average nitrogen oxide emissions reached “325 ± 90% of the respective emission limits”, and the average over individual measurement windows reached “up to a factor of 14”. Carbon dioxide on the road exceeded the laboratory figure by 21 ± 9 per cent.

What the same report found about everything else is the part that keeps this from being a story about a broken instrument. On-road nitrogen oxides from petrol vehicles, and carbon monoxide and hydrocarbons from both petrol and diesel, “generally stay within Euro 3-5 emission limits”. The laboratory number tracked reality tolerably for most pollutants and most vehicles. It came apart in one place, for one pollutant, on one fuel.

That report is dated 2011. It was published by the Joint Research Centre, which is part of the European Commission. It is a scientific and technical report, EUR 24697 EN.

Four years, then the scandal¶

In September 2015 the arrangement acquired a name and a criminal dimension. The European Parliament established a Committee of Inquiry into Emission Measurement in the Automotive Sector, which reported in April 2017 that the Commission and member states “were already aware, more than a decade ago” that diesel nitrogen oxide emissions in laboratory tests differed markedly from those on the road, and had failed to act appropriately to protect air quality and public health. It also found that no member state authority had suspected or investigated defeat devices, having performed only the standard type-approval tests.

Reading those two findings together gives the shape of the thing. The gap was measured, published and catalogued by the Commission’s own scientists. The defeat devices were not found, because nobody was looking anywhere except at the test.

Two failures that are not the same failure¶

A defeat device detects the test and changes the car’s behaviour during it. That is fraud, and it was prosecuted as fraud.

Optimising a calibration so that the car performs well over the cycle is not fraud. It is what a manufacturer is being asked to do by a regulation that defines emissions as what happens on a dynamometer. The distinction survives even where the same company did both, and only one half of it can be fixed by finding the wrongdoers.

Europe’s remedy addressed the second rather than the first. Commission Regulation (EU) 2016/427 of 10 March 2016 introduced a Real Driving Emissions procedure, putting portable measurement equipment on the road and making the road part of the question. That is not a penalty. It is an admission about what the previous question had been.

The clerk’s brief¶

April 2017: The inquiry reports¶

The European Parliament’s Committee of Inquiry into Emission Measurement in the Automotive Sector voted its final report, A8-0049/2017, on 4 April 2017. It found that the Commission and member states “were already aware, more than a decade ago” of the divergence between laboratory and road, and had failed to act appropriately to protect air quality and public health, and that no member state authority had suspected or investigated defeat devices, performing only standard type-approval tests. The clerks note that a committee of inquiry was needed to establish what had already been printed, and that the finding of not having looked is separate from the finding of not having known.

March 2016: The road joins the question¶

Commission Regulation (EU) 2016/427 of 10 March 2016 amended the Euro 6 implementing regulation to introduce a Real Driving Emissions procedure, using portable measurement equipment on public roads alongside the laboratory test. The clerks record the direction of the repair: nobody proposed abolishing the laboratory figure, which remains reproducible and comparable and useful, and the addition was of a second measurement rather than a correction to the first.

2011: The Commission measures its own question¶

The Joint Research Centre published EUR 24697 EN, “Analyzing on-road emissions of light-duty vehicles with Portable Emission Measurement Systems (PEMS)”, by Weiss and colleagues at Ispra. Twelve vehicles compliant with Euro 3 to 5 limits, four route types, and diesel nitrogen oxides at “325 ± 90% of the respective emission limits”, reaching “up to a factor of 14” over individual windows, while petrol nitrogen oxides and the carbon monoxide and hydrocarbon figures for both fuels “generally stay within Euro 3-5 emission limits”. Its own conclusion was that laboratory testing “fails to capture the wide range of potential on-road emissions”. The clerks note that the report opens by calling the laboratory procedure a source of “clear design criteria”, and that the phrase appears there as a description of the procedure rather than as a finding about it.

June 2007: The prohibition, written down¶

Regulation (EC) No 715/2007 of 20 June 2007 set the type-approval requirements for emissions from light passenger and commercial vehicles, and provided at Article 5(2) that “the use of defeat devices that reduce the effectiveness of emission control systems shall be prohibited”, subject to three narrow exceptions for engine protection, engine starting, and the test procedures themselves. The clerks note that the prohibition preceded the scandal by eight years, that the committee of inquiry found no member state authority had suspected or investigated a defeat device, and that a prohibition nobody tests against is a sentence in a regulation.

The number and the road¶

The dynamometer figures were accurate, reproducible and comparable, which is what they were built to be, and they were also the specification a vehicle was engineered against. The clerks’ standing assessment is that a regulatory test becomes a design specification without anybody gaming it, and that this specification was drafted by the regulator, published by the regulator’s own laboratory, and built to by the industry it was addressed to.