Yes, ballot questions in Colorado often are unnecessarily convoluted, often involving one huge sentence that no normal person easily can parse.
No, Initiative 234, which says “ballot questions shall be in plain language at no more than an eighth-grade reading level,” is not a good way to address the underlying problem.
Indeed, by failing to define what constitutes an eighth-grade reading level, the measure invites inane and costly legal battles. And defining reading levels is not as straightforward as you might at first suspect.
Imagine a lawsuit along the following lines: “The state’s ballot title registers on the Flesch-Kincaid Grade Level formula at a level of 8.0, but, if you properly add in a hyphen, thereby turning two short words into one longer word, the formula registers an 8.1. Because the constitutional language specifically says ‘no more than an eighth-grade reading level,’ the ballot title is unconstitutional. Na-na-na-boo-boo.”
Notably, the legislature just passed, and Governor Polis signed, House Bill 26-1320, requiring that the title board set titles that “are written using accessible language, which is plain language that is understood by the widest possible audience.” If the problem is that measure has dull teeth, the problem with Initiative 234 is that its teeth will be sharpened unpredictably by lawsuits.
A glance at reading scores
Who defines what is an eighth-grade reading level? That, as William Shakespeare so eloquently writes in his Flesch-Kincaid Grade Level 11 Hamlet, is the question.
I take it that “an eighth-grade reading level” does not refer to the reading abilities of Colorado’s actual eighth graders. As I pointed out last time, only 42.2% of eighth graders met or exceeded expectations on the state’s latest English test. If that were the standard, it’s unclear whether any ballot title could be dumbed-down enough to satisfy the requirements.
Presumably, then, we’re looking at some established formula for reading levels. But Initiative 234 doesn’t say that, nor does it say which standard we should use.
Should we use the Flesch-Kincaid formula, the Automated Readability Index, the Coleman-Liau Index, or the Gunning Fog Index? Or maybe we should use the SMOG standard (Simple Measure of Gobbledygook) just because that sounds cool?
Remember, if voters pass Initiative 234, this is exactly the sort of arcane minutia over which overeager lawyers itching to sue will salivate.
Let’s take Flesch-Kincaid as a common standard. Per Wikipedia, originally this was developed by Rudolf Flesch and J. Peter Kincaid back in the ’70s for the Navy. I’m sure any eighth grader would enjoy reading the original 49-page report.
The upshot is that the older Flesch formula was updated to the following: Grade Level = 0.39 (words/sentence) + 11.8 (syllables/word) – 15.59. That’s it. That’s the formula.
Why is it minus 15.59 and not minus 17.32 or plus 21? I didn’t read deeply enough into the report to figure that out. For our purposes, just accept that as the formula. And, in the background, imagine lawyers arguing the precise figures in court, which they certainly will do if this measure passes.
The first thing you should notice about this is that shorter sentences and shorter words (fewer syllables) yield lower grade levels. That’s the basic insight: short sentences, short words. “Write tight.”
Here is a very simple example. Consider the sentence, “The constitution is magnificent.” Because that short sentence contains two long words (plus two short ones), the Flesch-Kincaid score is 15.47. Of course the scale is not intended for use with single sentences; this is just to give you the flavor.
If we change the sentence to, “The constitution is great,” that drops the score to 6.62. On the other hand, if we try, “Our resplendent constitution is magnificent,” that gives us 17.04.
The Preamble to the Constitution (“We the People of the United States. . .”) weighs in at 25.8. The first sentence of Lincoln’s Gettysburg Address scores 15. The preamble to the Colorado constitution scores 29.2.
Initiative 234 fails its own test
Consider the full text of the proposed Constitutional amendment: “Ballot questions shall be in plain language at no more than an eighth-grade reading level. No statute may interfere with this provision by mandating language preceding, following, or in a ballot title of an initiative.”
If we count the hyphenated “eighth-grade” as a single word, that gives us 35 words, 58 syllables, and 2 sentences, for a Flesch-Kincaid score of 10.8.
Here’s something funny. If you instead remove the hyphen or just treat “eighth” and “grade” as two words rather than one, the score drops nearly half a grade level to 10.4. Either way, the language of Initiative 234 does not meet its own standard for dumbed-down language, at least according to Flesch-Kincaid. (Ballot titles are not the same as the underlying language, but still.)
Just for fun I tried to dumb down the language myself, but I only got down to a score of 8.8. ChatGPT, getting good mileage out of breaking the text down into three sentences, outdid me with a score of 4.5: “Ballot questions must use plain words. They must be no harder to read than eighth grade. No law may require words before, after, or in the ballot title of an initiative that would break this rule.”
But notice that “break this rule” doesn’t mean quite the same thing as “interfere with this provision.”
Also notice that I explicitly asked an AI system to optimize the language for a low grade score while keeping the language easy to understand.
I then asked ChatGPT to game the language along another dimension. It came up with a low, low score of only 2.5 but made “the syntax and references needlessly hard.” Here’s what it came up with: “A ballot question must use plain words. It must not be above grade eight. No act may bind a title to words if, but for those words, it would meet this rule. If it would not, this bar does not bar the act for that cause, though this rule still binds the title.”
So if you think that Initiative 234 will prevent people from gaming ballot titles, you’re a profoundly underclocked custodian of scarcely functional cognition. Or, if you wish, you can substitute for that flowery ChatGPT phrase the single-syllable term “dolt.”
Abusing grade scores
Advance Colorado, the organization behind the ballot measure (see the Sun), claims, “Colorado’s . . . ballot titles have consistently received grade-level readability scores between 21-43.”
Such claims seem to be coming from BallotPedia’s Flesch-Kincaid scores of various ballot measures. For example, BallotPedia ranks the ballot title for 2023’s Proposition II (tobacco taxes) a 43.
The first thing you should notice is that assigning something a grade level of 43 doesn’t make any damned sense, because there is no such grade level. Indeed, as experts in the field recognize, “The Flesch-Kincaid score is only valid between US grades 3 and 12 and cannot differentiate information that is easier or more difficult than these levels.”
The main problem with Prop II, in terms of its grade score, is that it’s one huge sentence. And it uses multi-syllabic words such as “cigarettes,” “tobacco,” and nicotine.” I asked ChatGPT to come up with a lower grade score, and it delivered a smooth-reading, six-sentence version at 7.7. (I won’t paste the whole thing.)
Then, at my goading, ChatGPT also came up with a “deliberately perverse version” by “exploiting short sentences, short words, vague pronouns, and unclear antecedents.” That version scored only 4.4. ChatGPT even helpfully summarized, “A Flesch-Kincaid ceiling cannot by itself guarantee clarity.” No kidding.
Other states’ efforts
Advance Colorado points out that other states have adopted comparable reforms. But the language of other measures is less troublesome, and even the best measures have invited legal wrangling. Let’s take a quick look.
I’m sure you’ll enjoy reading the 77-page opinion from 2025 involving a lawsuit over Arkansas’s grade-level requirement.
The Arkansas measure was a statutory change, not a constitutional one. And, unlike the Colorado proposal, it explicitly included a technical definition of “above eighth grade as determined by the Flesch-Kincaid Grade Level formula as it existed on January 1, 2025.”
That seems specific, right? Ah, but the definition does not say whether “above eighth grade” in this context means only up to and including 8.0 or also up to 8.99. Oops. So that’s part of what the resulting lawsuit was about (although the judge decided the matter on other grounds). The Colorado measure has the same ambiguity.
A New York measure suggested only that a board “shall use best efforts to score at an eighth grade reading level or below.” Because “best efforts” is a subjective standard, the inevitable result was a lawsuit. But the wiggle-room also gives judges more discretion. The language of Initiative 234 is far more inviting to lawsuits.
Rhode Island’s measure says a “question shall be stated in plain language reasonably calculated to be understood by persons with an eighth-grade reading level.” Here again the phrase “reasonably calculated” offers some discretion unavailable under 234. Notably, Rhode Island has only legislature-referred measures, not citizen-initiated ones.
BallotPedia helpfully summarizes similar measures in various states. The language of most of these other states looks similar to what Colorado already has on the books. There’s a reason other states have not adopted language like what Advance Colorado proposes.
Good intentions are not enough. Language matters. Unfortunately, the four-syllable word with which Initiative 234 will be most associated, if it passes, is “litigation.”
Ari Armstrong writes regularly for Complete Colorado and is the author of books about Ayn Rand, Harry Potter, and classical liberalism. He can be reached at ari at ariarmstrong dot com.
