In Defense of the Imperial Measurement System

The Hidden Logic in a Much Maligned System

Well, Germany appears to have come for America:

Dunking on the Imperial—more properly, US Costumary Unit—systems is easy and popular around the world. And yes, to outsiders, it does seem absolutely nonsensical. But it does have a lot of practicality baked in. No, come back! hear me out!

Distances

Of all the advantages of metric, the clearest to my mind is truly in distance. 5280 meters is 5.28 kilometers. The ease of conversion is indisputable and very clearly in metric’s favor. In each of the head-to-head comparisons, there is no great advantage to choosing one over the other. I slightly prefer kilometers for exercise like biking and running, but that’s because I can bike more kilometers than I can miles, not for any good reason. But the units divide well enough—even if the decimeter, dekameter, and hectometer are tragically underused in metric countries.

However, metric is impractical in one respect: it does not divide distances by thirds very well. A third of a kilometer is 333.33333333… meters while a third of a mile is 1760 feet. Dividing miles into thirds is not especially routine, but a third of a kilometer is not nearly as easily measured. This problem is much more practical at the smaller intervals; a third of a yard is a foot and a third of a foot is 4 inches.

If you’ve ever folded a letter, you know that folding into thirds is a relatively straightforward process. For yards and feet, where folding is commonplace, divisibility by 2 and 3 are strong virtues. Even if you cannot fold, say, a piece of wood, you can fold a string, mark it, and lay it out. On those fronts, Imperial is plainly superior. The 36 inches in a yard conversion allows for those 36 divisions to be found through four careful folds. While metric requires only 2 folds, folding by 5 is much harder.

Volumes

After waxing poetic about the virtues of thirds, it might seem like I am inclined to bash the Imperial volume system. After all, it is a mess of powers of 2; in the US customary units there are almost no other derivations and nearly every unit is some power of 2 in proportion to another.

What is the advantage of this? To find half of a volume of liquid you merely need to divide it between two identical containers so the heights are the same. As long as you can get reasonably similar sized containers, everyone at a transaction can be confident that they are looking at half. It is theoretically possible to do the same with fifths, as 10ths will ultimately demand, but you will give yourself a nosebleed trying.

The fluid once is twice the tablespoon. The gill (more commonly half cup) is 4 fl ounces, the cup is 8. A pint is 2 cups, and the quart is 2 pints. A gallon is 4 quarts. The only wrinkle in this whole system is the teaspoon, which is a third of a tablespoon.

It is worth also conceding the Imperial fluid ounce is 1/20 of a pint and is therefore silly. Americans were correct to rebel and establish a better system.

Weights

I will not belabor the point here as it follows volume fairly closely, but weights make a lot of sense as well. On a weighted scale, dividing by 10 is no easy task. But dividing by 2 is straightforward enough; just put equal amounts on either side. Therefore, the ounces system–16 to a pound–and the dracham—16 to an ounce—allows for many divisions of 2. The higher weights, e.g., the hundred-pound and long and short ton, are less reliant on this logic, though accurately weighing 100 lbs or 2000 lbs of anything in a pre-industrial society is relatively difficult, which probably explains the breakdown in the logic here.

Temperatures

Fans of the metric system love to tell me the virtues of the temperature system. After all, they tell me, zero-as-freezing makes perfect sense. I grant this. It is metric temperature’s only virtue for daily use. Yes, I said only.

But, I hear you say, what about 100 as boiling?

When was the last time you needed to know the boiling point of water for non-scientific use? On my electric kettle, water boils at the same temperature it shuts off; it could be e^2.8*τ for all I care. Virtually every temperature you bake at is above boiling by a fair margin and knowing that fact is useless to the process anyway. Most other culinary applications—like candy making—rely on temperatures that are essentially arbitrary in either system.

What you do use temperature for everyday is weather. I’ll grant Fahrenheit picked atrocious 0 and 100 points for his system, but he did luck into an excellent range. Choosing 0 through the process of mixing salt and ice water—again, stupid—had the effect of making 0 an exceptionally cold day in the temperate world. His choice of 100 was more arbitrary, but by sheer dumb luck he found the upper range of temperature in the temperate world. (And, indeed, even the tropics don’t routinely soar above that point, though the deserts do.)

The practical effect of this is that Fahrenheit is really good for describing weather. I have grown used to milliliters and kilometers, but Celsius just doesn’t work as well. I have taken those who have grown up with Celsius through the conversions pointing out that every 10 degrees you change your clothes and they have been immediately sold. If I say it is going to be in the 20s today in Fahrenheit, you know you need a winter coat with gloves and a hat on hand. The 20s in Celsius is everything from a nice spring day to full-on beach weather.

For daily use, Fahrenheit is truly superior.

Conclusion

To be honest, I’m actually for the US converting to metric, and believe we should have done so with the rest of the world. In an age of globalization and easily produced scientific instruments, the idiosyncrasies of the Imperial system are not worth preserving overall. But I do want it to get a proper send-off as a truly remarkable system for measurement that comes from an age of barter. Base 10 is clunky for everyday use and metric commits us to a system that requires official instruments.

And I think I’ve made the case for keeping Fahrenheit for weather—if I can learn the two systems, people can handle boiling water at 100 and also battling heat waves at 100.

But otherwise, this is a eulogy. It was quirky. It was practical. It was superseded by modern manufacturing. The US has clung to it for far too long. The time has come to wish it goodbye.

Meme I Hate: Boomers, Poverty, and Death

As the “Okay, Boomer” meme crests, the backlash against it has begun:

To my knowledge, no one has proved this with, you know, data. If I wanted to be really salty I could just end the post here. We’re done! Go home! No proof means you shouldn’t share the meme!

A Simple Model

But, let’s take a closer look. Some quick arithmetic can cast more doubt on this and the data doesn’t really back this up. Now, I want to start by not only conceding that the life-expectancy gap is real, but also candidly state that none of what follows is meant to suggest that reducing it shouldn’t be a policy goal. My sole purpose here is to say, “stop sharing this meme with no proof because it doesn’t really hold up”.

The gap between the very richest Americans and very poorest Americans is about 12 years on average. (It’s larger for men, partially due to the gendered nature of hard labor in the United States.) Conveniently, its pretty close to linear:

This is convenient because it allows for a stylized analysis. To avoid doing a linear regression, I used the 20th and 80th percentile points as samples and found the lines through each the male and female lines. Then I averaged the two. Because the errors are obviously pretty small*, we can rely on the following formula for our estimates:

Where E is a person’s life expectancy and P is their percentile in the income distribution. This was a lot of work to make a simple point. The difference between the bottom and the middle is 4 years and, likewise, the bottom and the top is 8*. Consider Boomers born in 1946, the very oldest boomers. Even the poorest boomers are still 5 years from life expectancy. The middle of the Boomers, born in 1955, are only just retiring next year and so significantly below the point where this effect would be most noticeable.

Again, we can stop here. A quick check shows that Boomers are young relative to life expectancy, even accounting for the effects of poverty.

Let’s Go All In

Now, people don’t automatically drop dead at their life expectancy—in fact, life expectancy has a pretty large standard deviation: 15 years. This means we can estimate the life expectancy gaps for each year after birth:

The purple curves are my estimates of what percentage of a birth cohort are alive at each year assuming deaths are normally distributed around the expected (average) death; the higher one, of course, is the richest 1% and the bottom one the poorest. The red curve is the difference. The gap gets the widest, 22% at 85 years old**.

Using averages across the 18 years of the Boomer Generation, we can show what the difference should be across time, here the green curve. The zero-year here is 1946, and 2019 is marked in purple:

So, the effect is pretty moderate, at 11%. But we’ve slipped one last assumption under the rug: lockstep class-voting analysis. The class-voting link is pretty moderate, with the poor voting only about 2:1 for Democrats and the rich almost perfectly reversing that. This probably evens out with the skew I introduced by assuming a normal distribution. We’re pretty far into the weeds at this point; the point here is that this effect isn’t that strong and so the meme is not that convincing.

But That’s Not All

That’s a neat model, no matter its shortcomings, but has an 11% change manifested since 2000? Not really:

While its true there has been a trend for the “lean republican” Boomers to become stauncher republicans, its not been effecting the conservative/liberal split very much. Further, a breakdown of the data for white voters indicates that its Gen Xers and younger Boomers who are driving the trend right; older Boomers—those most effected by the death gap—have been stable moderates for the last few years. I would concede the shortcomings of looking at data for white voters specifically, except it seems to be besides the point. Pew’s data is for all voters and shows the same consolidation to the far right among Gen Xers and Boomers. Unless the Gen Xers are dying off at rates equal to the oldest Boomers—and, no, that’s not happening—its probably safe to assume its mostly a change in public opinion among younger Boomers, not a function of deaths.

Worth noting that the rightward trend only shows up by about half in the Silent Generation; my model predicts a split of 21% for the Silent generation; Pew’s data shows a conservative drift since 2000 of only 10%. Given that class and voting are not lockstep, we can tentatively take that as suggestive that the effect identified in the meme is real, but small. Likewise, my model predicts a 9% change for the Boomers and the pew data finds 5%, further suggesting that if this is not coincidence, the effect is about half of my model.

Conclusion

So let’s take a step back. We’re looking at a 5% change in the Boomer generation’s voting patterns over a decade and a half. When looking at the claim “the reason why there are so many conservative Boomers is the poor ones die sooner”, its worth asking if 5% is really that big of a change. The numbers I’ve found seem to back the direction of the analysis, but when looking at the magnitude, I have to ask, is it that noticeable? If you took a small, random sample Boomers in 2000 and now to interact with, bias from random chance would be more noticeable than the partisan shift. And you’re kidding yourself if you think in the US you’re interacting with a random sample of Boomers given de facto race and class segregation.

So, sure, I’ve convinced myself the death gap effect is real, but all evidence suggests its would be hard to notice outside of statistical analysis.

*The biggest source of error in this equation if I were to use a linear regression would be the bottom 3%. As a consequence, my range is significantly smaller than the simple range given in my source. While the unique struggles of the bottom 3% deserve to be addressed, they add nearly 3 years to the range on their own and its representative to use them.

**By using a normal distribution, I’ve wayyyy overestimated the number of centenarians. But I think the basics of the point would hold up for a more left–skewed distribution, and, at any rate, we’ll see this model is not well borne out in the data no matter how you slice it.