Note: There are two somewhat significant updates at the bottom, just before the Appendix.
CAFE stands for Corporate Average Fuel Economy – the average fuel efficiency that is calculated annually for every manufacturer that sells cars or light trucks in the U.S. and compared to standards set by the National Highway Traffic Safety Administration, part of the Department of Transportation. (If you want to know more about how CAFE is measured, see the Appendix to this post.) Yesterday, President Obama proposed new, higher CAFE standards for models years up through 2016, by which point aggregate efficiency should reach 35.5 miles per gallon.
The typical conservative response to regulations like this is that they impose costs on the economy. In this case the main argument is that mandating higher fuel efficiency standards makes cars more expensive to produce; so car companies have to charge more for them; so fewer people will buy cars, and fewer people will be employed in the auto industry. I was planning to try to pre-empt this argument, but Keith Hennessey, former head of the National Economic Council under Bush II, beat me to the punch. His post summarizes some findings from a 900-page report produced by NHTSA in January 2008, when the Bush administration released the latest version of the CAFE standards. One of his main points, taken from that report, is that the Obama standards will cost 49,000 jobs. That’s relative to some baseline that I haven’t been able to identify, but it’s 38,000 jobs more than the Bush standards. The table is on page 586 of the long report; the Bush plan is “Optimized” and the Obama plan is “TC = TB.” Hennessey’s post has been picked up by Marginal Revolution (where I found it) and by The New York Times, so I decided I should stay up late and write a response.
Hennessey’s insight was to notice that the Obama plan looks similar to a plan analyzed under the Bush administration – TC = TB – and rejected in favor of Optimized. If you look at the first chart in his blog post you’ll see why he makes this guess, and it seems good to me. Based on that assumption, he can take the NHTSA’s analysis of the TC = TB plan and apply it to the Obama plan.
Optional interlude on “Optimized” and “TC = TB”
The Optimized plan is the one that, according to the analysis, maximizes net benefits to society – that is, benefits of the regulation minus costs of the regulation. The TC = TB plan is the one that gets the highest fuel efficiency while ensuring that the net benefits are not negative. The conceptual idea is that there is that there are diminishing marginal net benefits to higher fuel efficiency – you use less fuel and emit less carbon dioxide (“CO2″), but it gets more and more expensive to squeeze more efficiency out the engines, so the cars cost more and more.
I acknowledge this point, but I don’t put a lot of faith in it, because the calculation of benefits is highly dependent on your assumptions, such as the price of oil and the monetary value of reducing carbon emissions. In this study, the NHTSA used a range of $0 to $14 per metric ton of CO2, and used the midpoint ($7) in its estimates. The higher your estimate of the aggregate cost of CO2 emissions, the higher your optimal fuel efficiency standards will be.
Back to jobs
The argument that higher fuel efficiency standards reduces jobs goes like this. The diagram below shows the market before regulation. Auto companies will supply cars at some price that includes their cost plus the profit margin they need to pay off their debts and provide a return to their shareholders. (You could draw the supply curve with an upward slope and everything that follows would be essentially the same.) The more expensive the car, the fewer people will buy it. At equilibrium, the price is P0 and Q0 cars will get produced and sold.
Optional interlude on demand and supply curves: The curves show the quantity offered (supply) and bought (demand) at each price. Where they intersect is where the market should end up. This is incredibly confusing to people who know math but do not know economics, because the axes are backwards from the way they are in every other quantitative field that I know of. I believe this is the fault of some English guy named Marshall.)
Then regulation comes along, increasing costs for car makers. The Obama people have been using the number $1300, so let’s use that number. This acts something like a tax. The price charged goes from P0 to P1, where P1 = P0 + $1300. At the higher price, fewer people will buy the car, so quantity declines to Q1. Fewer people get cars (bad) and fewer cars get produced, meaning fewer jobs in the auto industry (bad).
However, there are two problems with this argument in its simple form. The first problem is that it isn’t the same car – now it’s more fuel efficient. That’s good, and it means that people will be willing to pay more for it. In other words, the demand curve shifts outward, so at equilibrium, quantity will be Q2, which is somewhere between Q0 and Q1. Q2 will still be less than Q0; otherwise, the free market would have come to that equilibrium by itself. (If people valued the increased fuel efficiency at $1300 or more, the industry would already have done it, at least according to a pure free-market argument.) So things are not as bad as in the picture above.
The other thing is that there is actually $1300 of additional stuff in the car, which someone has to build. So even if fewer vehicles are being made, they are not the same vehicles. If we take manufacturing cost as a proxy for labor inputs throughout the entire supply chain, then these vehicles require more labor than the pre-regulation vehicles. (You could say that you get the increased efficiency from more sophisticated components – but someone has to design and build those components, and if they were as easy to design and build as the less sophisticated components, then they wouldn’t cost any more.) So as a crude approximation, in the pre-regulation world the auto industry employs B + C people, while in the regulated world it employs A + B people. And there’s no way to tell from a conceptual drawing which is greater.
That’s the theory. But how do things work out in the real world? First I’ll look at the consumer side, then I’ll return to the production and jobs side.
The first thing to note is that the manufacturing cost of every vehicle will not go up by $1300 – that’s just an average. It’s also not as simple as saying that the manufacturing cost of gas guzzlers will go up more and the cost of fuel-efficient vehicles will go up less, because car companies can meet the standards any way they want, and increases anywhere in the fleet help. (They can’t just change their mix, as explained in the Appendix.) All other things being equal, you would expect more investment to go into the gas guzzlers, because the marginal returns to investment should be higher than in the Prius; but other things are not equal, like maybe the marginal bit of engine efficiency gets you more bang in the Prius than in a pickup truck, because it is already lighter and more aerodynamic.
On top of that, car companies have sophisticated pricing strategies that involve all sorts of cross-subsidization. For example, American manufacturers until recently sold small cars at small profit margins while selling large cars and trucks at huge margins. All other things being equal, you would expect them to discount the more efficient models and increase prices on the less efficient models with the same footprint (see Appendix), because that would have the net effect of increasing their CAFE, but again many things are unequal.
But let’s assume that prices will go up by some amount. Does that mean that fewer people will be able to afford cars, as implied by the diagrams above? Maybe. The flaw with all those diagrams is that they assume that there is only one car at one price, when in fact there are dozens of models, each with dozens of potential variations, all at different prices. For example, let’s say that in an unregulated world you would buy a top-of-the-line Toyota Camry XLE, with an MSRP of $25,575. Whoops, everything is now $1300 more expensive (let’s assume). Are you going to switch to mass transit? No, but you could switch to the Camry LE V6, which is $1300 less than the XLE. Or you could give up the navigation system, which has an MSRP of $1200. Or, if you want all the options, you could switch to a Ford Fusion, which is a little less expensive than a Camry, but which I hear is a very good car.
So would fewer new cars actually get sold? Maybe, at the margin, there are some people who are already buying the very cheapest new car out there (or the cheapest model in the class they need), and will be forced out. However, it is also possible that one or more manufacturers will come out with an even cheaper car to supply them – or simply keep a low-end model or two in their existing form, without additional manufacturing costs. In any case, the net effect should be smaller than in the theoretical one-model world, where there is no ability to switch into a cheaper substitute.
And, to take a normative stance for a minute, it’s not as if this is perpetrating some great injustice on the American consumer, like forcing people to take mass transit. There is a big used-car market in the U.S. (where my family has bought two of the three cars we have ever owned), and, believe it or not, there is no Constitutional right to a new car. Theoretically, increased prices for new cars could ripple into the used car market – eventually – but you can still substitute into slightly cheaper models. At the bottom end of the spectrum, people who can barely afford any car are looking at things like my 99 Chevy Prizm – and I doubt that new CAFE standards are going to boost its resale value by more than a few cents.
So, I think I’ve outlined a number of reasons why job losses should not be as great as in a crude “$1300 tax” analogy in a world where only one car model exists: demand for cars will go up because they are better; there is more stuff in the cars that has to be built; and people can substitute into different new cars instead of being forced out of the new car market. The fact remains that the NHTSA report cited by Hennessey estimates 49,000 job losses relative to the baseline scenario which, I assume, is the world prior to the 2008 Bush Administration change to CAFE standards for model years 2011-15. So we need to look at what’s behind that number.
In the main report, the entire discussion is on pages 584-86, and this is what we have on methodology:
The calculations assume that compliance costs are passed onto consumers in the form of higher prices. These higher vehicle prices (net of the benefits of added fuel savings and added resale value) lead to reduced demand for vehicles. Estimates of sales losses are made using the price changes and the elasticity of demand for new vehicles (-1.0). Losses in sales are translated into losses in jobs by dividing through by the average number of vehicles produced per full time jobs in the automotive industry (approximately 10.5 vehicles per job).
So it looks like they did take into account the fact that the vehicles are worth more. Another thing to note is that the estimate covers the entire auto industry, so the 49,000 jobs number includes overseas jobs; the number of domestic jobs would be considerably lower. But it’s not clear whether there is one model with a single type of car, or a more sophisticated model.
The main report cites Chapter VII of the Final Regulatory Impact Analysis (FRIA), which I cannot find on the NHTSA web site. I am looking for the FRIA for the Final Rule for model years 2011-15, which became effective in January 2008. I can find something called the Preliminary Regulatory Impact Analysis for those model years, which is actually dated April 2008, and is probably pretty close.
In that report, the relevant section, “The Impact of Higher Prices on Sales,” begins on page 259. There is a long explanation of how they estimated the increased value of a more fuel-efficient car, which looks pretty rigorous. But this is a crucial paragraph:
A sample calculation for Ford passenger cars under the Optimized 7% alternative in MY 2011 is an estimated retail price increase of $782, which is multiplied by 0.911 [to reflect the increased resale value of the car, less some increased costs of ownership] to get a residual price increase of $712. The estimated fuel savings over the 5 years of $281 at a 3 percent discount rate results in a net cost to consumers of $431. Comparing that to the $21,821 average price is 2.39 percent price increase. Ford sales were estimated to be about 1,300,000 passenger cars for MY 2011. With a price elasticity of –1.0, a 2.39 percent increase in sales could result in an estimated loss in sales of 3,104 passenger cars at a 3 percent discount rate.
So the model assumes a world where every manufacturer makes exactly one passenger car and one light truck and applies a price elasticity of -1.0. Now, it’s possible that that price elasticity was estimated using a similarly aggregated model of the world, so perhaps this approach is accurate. But I’m also worried about the $782 estimate for retail price increases. The methodology for estimating manufacturing cost increases is complicated, to say the least. It is based on estimates of the cost, effectiveness, and availability of specific new technologies (pp. 88-93). Then these are applied to data from the manufacturers about their production plans (see pages 126-28). As far as I can tell (I admit, I didn’t read every word), the algorithm takes these car company plans as a given, meaning that it does not allow them to shift their production plans in response to changing CAFE standards. This seems unrealistic to me.
While I’m skeptical, I don’t have any strong evidence that the estimated sales losses are too high. However, there is a bigger problem on page 271:
There are three potential areas of employment that fuel economy standards could impact. The first is the hiring of additional engineers by automobile companies and their suppliers to do research and development and testing on new technologies to determine their capabilities, durability, platform introduction, etc. The agency does not anticipate a huge number of incremental jobs in the engineering field. Often people would be diverted from one area to another and the incremental number of jobs might be a few thousand.
The second area is the impact that new technologies would have on the production line. Again, we don’t anticipate a large number of incremental workers, as for the most part you are replacing one engine with another or one transmission with another. In some instances the technology is more complex, requiring more parts and there would be a small increase in the number of production employees, but we don’t anticipate a large change.
[The third is lower sales.]
It seems to me that they are ignoring the fact that more complicated, more expensive cars with more stuff in them require more labor to build. Essentially, this study is saying that it will cost more to build more fuel-efficient cars because they require new technologies that must be developed and manufactured at a higher cost than the old technologies – yet those new technologies and components will not employ any more people than the old technologies. If they don’t employ any more people – anywhere along the supply chain – why are they more expensive? I doubt that the new technologies use more iron or copper than the old ones; if the increased value doesn’t come from raw materials, it must be added by people, somewhere along the line. A little might go to intellectual property licenses, but that can’t amount to all of the incremental cost.
Anyway, something seems deeply wrong to me here. In short, this approach denies that box B on my last diagram has any impact on employment.
In the end, I’m willing to acknowledge that higher fuel efficiency standards should mean slightly lower new car sales, which could lower auto employment a little. But the 49,000 number is definitely high for at least two reasons: (a) it ignores the additional jobs required to build more complicated, more expensive cars; and (b) if you’re looking at this from a U.S. standpoint, it includes foreign automakers. And it may be high because it doesn’t fully reflect substitution between car models, either by manufacturers or by consumers.
I have something to say about the environmental impact as well, but that will have to wait for tomorrow, as this is already my longest blog post ever.
Update: Lying in bed last night, I changed my mind. I’m not willing to concede that higher fuel efficiency standards mean fewer auto industry jobs; they might, but it’s not certain.
The elasticity of a product is the amount that purchases of that product will change in relation to changes in its price. We say that gas and cigarettes have low elasticity: double the price of gas, and people will still buy almost as much of it; the same goes for cigarettes. Elasticity is expressed as the ratio of the proportional change in sales to the proportional change in price. So if the elasticity of cars is -1.0, that means for a 3% increase in price, you will get 3% fewer cars sold.
The implication, which I missed last night, is that almost the exact same amount of money gets spent on new cars (1.03 * 0.97 = 0.999) – there’s just more money per car. If the price goes up because manufacturers are able to raise prices, keeping the cars the same, then more money (per car) goes to their profits, so employment can go down. The same thing is true with a tax, where more money goes to the government. If the price goes up because the cost of steel goes up, then more money goes to steelmakers, so employment in the auto industry can go down.
In this case, though, the price increases are because the cars have more car stuff in them, increasing their manufacturing costs. So the amount of money being paid into the auto industry is the same and the profits are the same, and money going to raw inputs is the same. So auto industry employment can only go down if labor’s share of costs goes down, or if per-worker compensation goes up. In the absence of a reason why one of those should be happening, I don’t see why employment would go down.
Also, thanks to Oliver, I fixed a typo in the first paragraph – 2006 should be 2016.
Update 2: RJ raised the safety question in this comment. Dominic J, winstongator, odograph, and StatsGuy responded in replies (directly below the original comment), saving me the trouble of doing the research. Thanks.
Appendix: Interesting Facts About CAFE
There are many things I didn’t know about CAFE until I did the research for this post. Here are three that may be interesting.
1. CAFE does not fall into the miles-per-gallon trap. If you take two cars, one that gets 10 mpg and one that gets 20 mpg, and drive them the same distance, you will not get an overall average mpg of 15; you will get 13.3. Try it yourself if you don’t believe me. You do not have this problem if you use gallons per mile instead; in that case your two cars would use 0.1 and 0.05 gallons per mile, and on average they would use 0.075 gallons per mile, which is exactly what you would expect.
This is the formula that CAFE uses:
where N is the total number of cars, Ni is the number of model i, and MPGi is the MPG of model i. It says it uses MPG, but if you stare at it long enough you’ll work out that the denominator of the fraction is calculating how much gas it would take to drive every car one mile, which is a gallons-per-mile calculation, and everything works out.
2. There is a different standard for each vehicle “footprint,” which Hennessey helpfully defines as “the shadow made by the vehicle when the sun is directly overhead.” So if you have bigger cars, you are allowed to have lower gas mileage. One implication is that you can’t meet higher CAFE standards simply by dropping SUVs and increasing sales of small cars, since your aggregate standard (that you have to meet) is a weighted average of the standards for each of your models – weighted by the number of each model that you sell.
3. CAFE compliance is based on actual vehicle sales, not predicted sales. I was a little suspicious about this.
By James Kwak