Well, at least I’m pretty sure we should.
2020 is another census year, which means there will be some electoral shakeups around the country as fast-growing states gain members of Congress and slower-growing/declining states lose representatives. For almost 100 years now, we’ve had this fixed-pie system at the national level, where states could lose seats in Congress if they aren’t keeping up. But that wasn’t always the case. When the US constitution was adopted, the House of Representatives was meant to have 1 member for every 30,000 people. For today’s population, that would come out to more than 11,000 members of Congress. Sure, it might mean that we have 25 AOCs, but it would also mean 25 Jim Jordans! No thanks.
Clearly, capping the number of Congressional Representatives at some number is necessary. 11,000 members of Congress is probably untenable, and fivethirtyeight.com’s name would seem so quaint. There have been some calls to increase the size of the house, and there is some good research out there about what the optimal size of the legislature should be. TLDR: the number of representatives in the lower house should be the cube root of the entire population, so in the US, the House should have around 688 members.
But what about state legislatures? So much policy happens at the state level: healthcare, transportation, zoning, criminal justice, minimum wage as well as culture war issues like abortion, gun control, and drug laws. It seems like we should also consider what the right levels of representation is for our state legislatures is too.
I played around with the data, and found that some states have really high levels of representations (North Dakota and Vermont, but whoa New Hampshire!), while others have a lot of people per representative (California in particular, but also Texas and Florida). Others are just weird (why only one house Nebraska, and do you really think the state legislator is a non-partisan position?). Here in Ohio, I think there’s a decent argument for a larger legislative body but before we get ahead of ourselves, let’s dig in!
Comparing State Legislatures
To compare levels of representation across the states, I pulled together the populations and sizes of state legislatures for every state. Somewhat arbitrarily, I decided to add the number of representatives upper and lower houses and divide by the population. In Ohio’s case, we have 99 representatives and 33 senators for a grand total of 132 legislators. The population is about 11 million, so I took 11 million, divided it by 132, and found that there are about 87,000 people per legislator in the buckeye state. This ratio does not exactly showing the number of people each member of the legislature represents because of overlaps in the the upper and lower districts, but it does give a rough idea of how many people a member of the legislature represents. It also takes into account variations in the ratios of upper house members to lower house members across states. Some states have large Senates relative to their Houses, and vice versa. Nebraska’s Senate is infinitely larger than its house (Nebraska is unicameral – that’s supposed to be a math joke).
Okay – here’s the comparison across all the states:
Clearly, there’s a pretty broad range, with a typical California legislator representing one hundred times more people than a typical New Hampshire rep. This is partly because New Hampshire has a wildly large lower house – 400 members, almost double the next largest (Pennsylvania’s with 203 members), and a small population. California has a relatively small number of representatives, only 120 in both houses, 12 fewer seats than Ohio, and a large population, almost 4 times larger than Ohio.
With a somewhat smaller number of representatives and a fairly large population, Ohio is the 5th least representative state, according to this ratio. The buckeye state falls behind New York, Florida, Texas, and California, the four most populous states in the country. Here’s the data represented as a very rough map:
From the map, it doesn’t look like geography plays a role sizes of legislatures. While the smallest legislatures tend to be in New England and the West, there are large and small legislatures in all parts of the US.
Instead, states with higher populations tend to have lower levels of representation, which makes sense. Even then, the relationship isn’t so clear. Even taking away the weird outlier states: New Hampshire, California, and Nebraska, the correlation is relatively small (r-squared=.23), which really only tells us that some states have really large legislatures while others have really small ones. Here’s what the scatter plot looks like:
So, Is There an Optimal Size?
The short answer is probably, but it’s hard to determine what exactly is optimal.
Initially, I thought that there might be a link between the size of the legislature and the level of partisan mismatch at the state level. That is, I expected that states with larger legislatures would have a mix of representatives that roughly matches the overall party affiliation at the state level. In other words, more representatives would mean that the legislature more closely matches the state-wide party affiliation rates. To check this, I took the state-by-state votes for Hillary Clinton and Donald Trump from the 2016 presidential race and subtracted the percentage of votes from the the partisan breakdown percentage in the within the legislature. Here’s how I calculated it:
y = (c-d)+(t-r)
Where:
y = Overall Mismatch
c = % voting for Clinton
d = Democrats' share of seats in the state legislature
t = % voting for Trump
r = Republicans' share of seats in the state legislature
It turns out, this was kind of a blind alley. There is a small reduction in the mismatch as the legislature grows, but the relationship is pretty small. Even when you throw out the weird states the r-squared is only .11 indicating that there’s not a strong linear relationship, for non-stats nerds, it’s not a slam dunk.
So, until someone does some more sophisticated research, we’re left with gut responses and not enough evidence to warrant making big changes. Nonetheless, I think the cube-root rule (noted above and in this research) is interesting, so, for fun, I charted out the difference between the actual size of the lower house and the predicted size based on the cube root of the population:
Based on the cube-root rule, California should add 254 seats to its upper house while New Hampshire should remove 290. 45 of the 50 states have legislatures that are too small according to the cube root rule. Apparently Montana is killing it – their lower house matches the expected size exactly – nice work, Montana. Ohio’s House is short by 127 seats, almost double the size of both houses of the Ohio General Assembly!
A Bigger General Assembly, Maybe
If we grew Ohio’s House to match the cube root rule, we’d end up with 225 seats in the House and 75 in the Senate (assuming we keep the same 3-reps-for-every-senator rule). Each representative would be responsible for about 52,000 Ohioans and each Senator for about 156,000, instead of the current 118,000 and 354,000 respectively. It would move Ohio from the 5th least representative state to 20th.
Such a change could have a big effect on our relationship with our representatives. There would be a massive reduction in the number of constituents that each representative or senator represent. Check it out:
| Currently | “Cube-Root” | Change | |
| # Reps | 99 | 225 | +126 |
| # Senators | 33 | 75 | +42 |
| People per Rep | 118,081 | 51,956 | -66,125 |
| People per Senator | 354,242 | 155,867 | -198,376 |
| People per legislator | 88,561 | 38,967 | -49,594 |
For cities, in particular, the level of representation would be very different. Columbus would have 17 representatives and at least 5 senators, which is over 3 times the number of seats on the Columbus City Council (but like, what the heck, Columbus?). Cleveland would have 7 reps and 2 senators, while cities like Toledo would have 5 representatives and at least one senator. Less populated areas would benefit as well: cities like Lorain, Kettering, and Springfield would each have their own representative and rural areas like Knox County would have its own representative instead of being bunched into a district with Delaware County and parts of Westerville and Columbus.
The data I played around with didn’t seem to suggest that there’s a strong relationship between legislature size and the overall mismatch of representation. However, it seems that it would be harder to gerrymander smaller districts, especially if we required that the districts follow city and county boundaries as much as possible. It’s definitely worth exploring some more (maybe another blog post!).
Regardless, I think there is a “just right” space for the size of the legislature – it is possible to be too small or too large. For me, the key question is, “is Ohio in the ‘just right’ space, or are we under- or over-represented?”
I’m honestly not 100% sure what the right answer is, but I don’t think we’re over-represented. We currently rank pretty low on the level of representation among the states, and over 100,000 people per representative seems high. With that many people to represent, I think it’d be easy for a legislator to become totally out of touch with their constituents (which might be what’s going on).
And so, I’m left without strong data but with a strong gut feeling. I think a strong democracy requires a close connection between the represented and the representatives. That’s why if I had to make a call on it, I would opt for more representatives in smaller districts. If we implemented the cube-root rule and increased the legislature to 225 reps, state reps would live fairly close to their constituents (except in smaller counties). In some cases, constituents would see their representative around town, and it would be much easier to interact with them on important issues. It would mean that local representatives would represent folks in the state Capitol. For better or worse, more voices would be heard. I think it’d be better.
I’d be curious to hear what you think though. Hit me up if you have strong opinions about this (or if you don’t) or know of some good research. And if you’re interested in my code, check it out on git.