Labor Day strikes me as being the ideal time to talk about a topic that may be of greatest interest to fellow research nerds: What kind of information is good enough for American automotive history?
I started to percolate on this question after reading a post by Aaron Severson (2025) at the Patreon arm of Ate Up With Motor. He discussed the challenges of trying to get access to primary sources of data on automotive sales and production levels.
Severson (2025) quite rightly noted that secondary sources of data, such as the Standard Catalog series, “are often riddled with errors, a lot of them simply typographical.” He is also self-aware enough to acknowledge that he would prefer to make his own “mistakes rather than just inheriting them from others.”
How I have built my production databases
I agree with Severson’s general sensibility. That’s why I have not drawn upon the database posted by Curbside Classic last year (Niedermeyer, 2024). There are so many ways to go wrong in working with this data that I would rather do it myself. For starters, his citation appears to conflate the title of two different books: “Standard Encyclopedia of American Cars.”
In building my own databases I have tended to draw upon multiple sources, particularly when some data doesn’t look right or is incomplete. In general, I have found the most questionable data in the Standard Catalog series (e.g., Gunnell, 2002). However, I have occasionally had issues with data from books and magazines that draw production data from the auto editors of Consumer Guide. This includes their reference book, The Encyclopedia of American Cars (1993, 2006).
When building a data table I tend to look at both of those sources as well as J. “Kelly” Flory’s American Cars series (2004, 2009, 2013). When none of the data is fully satisfying I look for other sources such as brand-specific books, websites and Wikipedia pages.
I try very hard to be methodologically consistent when building a graph or discussing data within the text, such as by not knowingly mixing calendar-year and model-year data. I mostly use the latter for consistency’s sake but will sometimes draw upon calendar-year data if it answers a specific question better. In those instances, I tell the reader that I have made the switch either in the text or in the “Notes” section at the bottom of an article.
Yes, but why not use more primary data?
It would be better if I could get access to primary sources, such as Automotive News market data reports. I draw upon those when I have them, but more often than not my stash of past issues doesn’t go back far enough. To make matters worse, I live in a relatively small town lacking in a decent research library.
I have thought about taking the time to travel to a major regional university library and camping out there until I compiled a decent amount of data. Alas, I have thus far not gotten around to doing it. A big reason why is that the day-to-day necessities of running a website can overshadow “nice-to-have” projects.
My current databases are about as good as any of the major reference books, so I have thus far concentrated on fine tuning them rather than rebuilding them with primary data. For example, when reposting an article, I will spot check my figures in the database. Sometimes I find an error or a newer source of information, so will update the text and any graphs.
How to deal with major reference books displaying data discrepancies? Beyond comparing sources and picking what appears to be the most accurate figures, I try to lower expectations about the precision of the data. This is why in the text I usually round off the numbers to the nearest thousand (but use exact numbers in my spreadsheet calculations). I also post line graphs rather than raw data tables like in Collectible Automobile.
Individuals aren’t going to solve the problem
I admire Severson’s diligence in tracking down primary data sources, but I suspect that this is a losing cause when done by individual researchers. Only a university or a major nonprofit organization would arguably have the capacity to create and maintain a large database with an adequate level of methodological rigor.
This is not a new idea — I wrote about it two years ago when discussing in somewhat greater detail this topic (go here). At that point I expressed skepticism that a for-profit entity was the right venue for such a project — and I am even more skeptical now.
Who could do this? I previously mentioned the Antique Automobile Club of America. In theory, other groups such as the Society of Automobile Historians could also take this on, but I fear that volunteer-based efforts could struggle to launch and maintain the project. This one strikes me as more the province of paid staff with quantitative research skills.
Here we have yet another reason why the U.S. needs at least one university to take a greater interest in automotive history. As a case in point, once upon a time the University of Michigan at Dearborn co-sponsored the Automobile in American Life and Society. Alas, that website hasn’t been updated in years, which suggests that it no longer has any champions within the university.
Academia is asleep at the wheel. Why is that?
Share your reactions to this post with a comment below or a note to the editor.
RE:SOURCES
- Auto editors of Consumer Guide; 1993, 2006. Encyclopedia of American Cars. Publications International, Lincolnwood, IL.
- Flory, J. “Kelly” Jr.; 2004. American Cars, 1960-1972. McFarland & Co., Inc.
- ——; 2009. American Cars, 1946-1959: Every Model, Year by Year. McFarland & Co.
- ——; 2013. American Cars, 1973-1980. McFarland & Co., Inc.
- Gunnell, John; 2002. Standard Catalog of American Cars, 1946-1975. Revised Fourth Ed. Krause Publications, Iola, WI.
- Niedermeyer, Paul; 2024. “US Car Sales And Market Share By Brand And Model Category, 1946 – 1975 – A CC Exclusive Database.” Curbside Classic. Posted Jan. 5.
- Severson, Aaron; 2025. “All the Data in the World Except the Part You Want.” Ate Up With Motor. Posted Aug. 24.
ADVERTISING & BROCHURES:
- ads.aacalibrary.org (Antique Automobile Club of America): Chevrolet (1952, 1953, 1958, 1963, 1965, 1969, 1977)
It’s always going to be a problem. What is being reported? Model year? Calendar year? Any automakers use some fiscal year not a calendar year? And when is the car counted? Out the factory door? Delivered to dealer? Sold? Each end point tells different information that could be important. Let’s toss in some of the independents who would retitle last year’s model to this year’s. By the mid-60s the US/Canada auto industry was pretty much a single unit. Good luck getting an exact count. For historical purposes you want to get the most precise figures, but define what you are working with.
Agreed. I strive to be as accurate as possible while recognizing that the data is arguably best used to show general patterns.
In addition to methodological questions about of how to count cars, there are also questions surrounding the best way to handle the available data. For example, I type into my spreadsheets exact production numbers rather than rounding up because the latter can meaningfully distort total figures once you add up all of the body styles, trim levels and models.
Do you include information for your database from Marti reports (Ford) and Glover (Mopar)?
I haven’t but will take a closer look at them; thank you for the referral.
My main focus has been comparative production levels, e.g., how did one nameplate do against its rivals for a given number of years? Unless I find serious problems with a specific dataset (such as from the auto editors of Consumer Guide), my tendency is to stick to their numbers because they are presumably “internally consistent” — that is, the same methodology is used in tabulating them. Once you start mixing and matching numbers there is a heightened chance of comparing apples to oranges.
Of course, there are exceptions to this rule, such as when a book obviously switches from model-year to calendar-year figures. I steer clear of those datasets. I found that to be the case with a Consumer Guide book titled Over 100 years: The American Auto.