A while ago, I looked into the carbon footprint of manufacturing a car, and found the GREET model (short for "Greenhouse gases, Regulated Emissions, and Energy use in Transportation"), a lifecycle analysis developed by a team at Argonne National Lab.
For a "vanilla" car weighing 3330 lbs (think a Honda Accord sedan), they arrive at 8.5 tonnes CO2 for the Vehicle-cycle Energy. That's 8500 kg or about 18,800 lbs of carbon dioxide (or CO2-equivalent greenhouse gas).
Driving this vanilla car for 160,000 miles at an efficiency of 24.8mpg (both typical numbers for cars in the US) emits 68.4 tonnes CO2 from burning the gasoline. That's a "well-to-wheel" number, and includes the energy required to pump the oil out of the ground, refine it, and transport it to your local gas station.
So in its lifetime this vanilla car causes 76.9 tonnes of CO2 emissions, of which 8.5 tonnes, or 11%, accounts for its manufacture. So much for the argument to drive your old gas guzzler until it falls apart.
But wait: I have seen other numbers. And those numbers, quoted often by European writers, suggest that building the car causes up to half of its total lifetime carbon emissions.
Half is a lot more than 11%. Half suggests that the planet is better off if you did keep your old guzzler until it falls apart. Okay, European cars are more fuel efficient, on average, than American cars - but surely not enough to account for that large a difference.
I was puzzling over the discrepancy until recently I read Mike Berners-Lee's book, "How bad are bananas? - The carbon footprint of everything". It's a cool book, very readable, and covers everything, from sending a text message (0.014g CO2-e), an ice cream from an ice cream truck (500g CO2-e), to a computer (200-800 kg CO2-e), to the whole world (50 billion tonnes per year, and rising). Very useful for setting out to slash your personal carbon footprint.
One of the items in the book is the carbon footprint of making a car, and indeed in rare cases that can be half of its total lifetime footprint. If you don't want to read the book you'd be missing a bunch of interesting stuff, but you can still read a re-cap of the automotive bit in this Guardian article.
Berners-Lee argues that doing a lifecycle analysis by looking at the underlying processes tends to underestimate the total carbon footprint, simply because it is too easy to overlook some contribution, no matter how painstakingly you go about the analysis - and obviously, the GREET model is nothing if not painstaking.
Instead of building the total carbon count up from every nut and bolt that goes into the car, Berners-Lee takes the macro, top-down, approach and asks, how much energy does it take to produce a dollar's worth of car? This is known as an input-output analysis of the carbon intensity of a manufacturing process.
Such an analysis makes for easier estimates (well: a bit easier). I don't understand the underlying economic math, but I imagine the estimate includes such things as the electricity needed to drive the conveyor belts at the assembly line, the pencils (OK, and computers) used by the development engineers, the power for the elevator carrying autoworkers to the cafeteria, the marketing, the private jets used by the executives as they go to London / Washington / wherever to lobby the government on behalf of the auto industry, and everything else that makes the auto industry go.
Coming at it from this angle, Berners-Lee arrives at an estimate of 720 kg CO2-e per £1000 worth of car.
See, Berners-Lee is a Brit, so he measures carbon emissions in kilograms, and money in pound sterling.
I am actually not being flippant: his estimates are derived from data of the UK's Office of National Statistics. In the American edition of his book he translates a car's carbon intensity into "450 kg CO2 for every $1000 of the price of a new car", by doing a straightforward conversion of the pound sterling to the dollar. But this translation is dicey because the OTR (on-the-road) price of cars in the UK is about 50-60% higher than in the US, once you add various sales and vehicle taxes. More on that below.
Following this recipe, Berners-Lee considers, among other cars, a Land Rover SUV: the high-end Discovery SDV6 3.0L. The UK price is £57,750 (at today's rate US$90,950), implying a footprint of 41.6 tonnes CO2. ( In America, this SUV is known as the LR4 and costs at most $59,220 even with a 5.0L V8 engine; see what I mean about the pricing?).
Driving it for 100,000 miles at 23 mpg causes emissions of 46.6 tonnes CO2. Assuming you scrap it after that, then yes, the manufacturing accounts for 43%, nearly half, of its total lifetime carbon footprint.
But how many cars get scrapped after only 100,000 these days? The US lifetime average is about 160,000 miles. Even in Europe it's 130,000 miles, and nearly twice that for diesel cars.
So let's look at a more typical vehicle.
As a concrete example, consider the six-seater Mazda 5, with kerb weight 3417 lbs (very close to GREET's "vanilla" car of 3330 lbs), and fuel efficiency 26 mpg. Under the GREET model, there would be 8.5 tonnes CO2 in its making. Its MSRP in the US is $19,600. At current exchange rates, that is approximately its price in pound sterling; according to Berners-Lee's analysis its manufacture would cause 0.720 * 19,600 = 14.1 tonnes CO2. In other words, 1.7 times the GREET estimate. That sounds like a big difference, but that's just the reliability of these very difficult estimates.
Driving this car for its US average lifetime of 160,000 miles would cause total lifecycle emissions of 80.3 tonnes CO2, of which 18% attributable to its manufacture, according to Berners-Lee's input-output analysis. Remember in the GREET process-based analysis, manufacture is 11% of total emissions.
For my own purposes, I'm going to treat this spread as an error bar, albeit a large one, in the estimated carbon emissions to build a car: building a Mazda 5 causes as least 8.5 tonnes CO2 emissions, and as much as 14.1 tonnes. My personal guess is that Berners-Lee's higher estimate is closer to the actual total carbon emissions than the GREET estimate.
But even using the high estimate, the manufacturing emissions is only half the total lifetime emissions in the case of conspicuous consumption: e.g. when you buy a large luxury car and scrap it after driving it a scandalously small total distance like 100,000 miles.
I'm starting to see how imposing a carbon tax on goods at the point of sale, such a powerful and simple feedback mechanism in principle, can be a nightmare to implement.