Last week (or so), I proposed a National Electoral Institute, or NEI. The short version of that proposal: The NEI is essentially a National Transportation Safety Board for elections. I promised to discuss the composition of such an organization as well as barriers to its creation and downsides, so here ’tis. This post is perhaps shorter than it could be, but I figured I better get off my butt and finish up before the election….

Staffing and Composition

So who would I hire to my hypothetical NEI? In addition to the usual lawyers, I’d consider:

  • Human Factors Engineers. Interface design? Yep, there’s science on it, and quite a bit. Anyone who has designed a user interface (me, for instance) knows the difference between something done hastily and something done with careful attention to detail. It is NOT easy. Cognitive scientists and human factors engineers cooperate to optimally design airplane cockpits and car dashboards, for instance. They can optimally design ballots to be readable to people with vision issues, such as the increasing number of elderly in the country, to avoid confusion, and otherwise improve the ballots to avoid the dreaded Butterfly Ballot Fiasco, or the numerous other fiascos we never heard about but doubtless exist out there.
  • Computer Scientists. One of the big holes in the system is computerized voting. There are all sorts of good reasons for computerized voting—cost, rapid counting, etc.—but getting the security right is far from trivial. Specialists in networking, formal methods and security would all play an important part. For a good example, read about the use of invisible ink on optical scan and voter-based fraud detection here.
  • Mathematicians and statisticians can help design innovative fraud detection schemes. For instance, the invisible ink fraud detection scheme I linked to previous is designed to preserve secret balloting while allowing a voter to check his or her own ballot. By having Joe the Voter (maybe even Joe the Plumber!) check his own ballot the potential for fraud through altering votes is dramatically reduced. Now it turns out that the vast majority of people won’t check, or wouldn’t notice even if they did check, but a clever statistician has shown that only about 1% of people need to check before widespread fraud will be noticed with very high probability. That’s the sort of thing that statisticians think about but most lawyers simply wouldn’t.
  • Political scientists with expertise in polling, geographical statistics, districting and voting rules would also be very helpful.
  • Oversight It’s really, really important to have the oversight of something like this be insulated from the political process or we’ll just be back to Gonzo land again soon enough. Thus it would make sense to have an oversight board with partisan representation, but the composition of the board needs to be people who are insulated. We can take a page from Iowa, where Congressional redistricting is done by a board of retired judges, not the state legislature.

Many of these people are already working on these problems out in academia or industry, e.g., Michael Herron at Dartmouth or Ed Felten at CMU, and that’s good. However, having a group that has institutional memory and, more importantly, a pot of hard money to give to researchers and to pay for “tiger teams” to handle problems when things go really wrong, would be very useful.


Well one of the big barriers to an adoption of this sort of thing is, of course, the Rovian political operator types (of all parties), who have a big incentive in having the system be manipulable. They’re not going to like having oversight and standards. And anyone proposing such a system should be prepared for a big fight from them. But I say screw them as they are surely screwing us.

However, there is a real risk of such an organization becoming a stultified bureaucracy, which is why limiting its enforcement powers would probably be wise (the “who guards the guardians?” question), and this factors into one of the giant barriers in the way… the US Constitution, which expressly leaves running elections up to the states. Massive federalization would require a constitutional amendment, which probably falls in the “ain’t gonna happen” column and I’m not sure I think it should happen. However, I suspect that many states would be very happy indeed simply not to have to mess with things if there was a good, low cost set of options available to them. If there were a list of approved systems certified by my hypothetical NEI, many states might well adopt. Even more, if the Federal government were to fund use of such voting systems via grants, this would shift quite quickly.

I don’t think I need to remind readers that (a) there is an election coming up real soon now and (b) the good old U-S-of-A has not had a good run of elections that are both clean and appear to be clean of late, as the various controversies of the 2000 and 2004 elections show. As a reminder:

  • Diebold Corporation, maker of touch screen voting systems seems to have a cozy relationship with Republican Party operatives and, far more importantly, has held the code for its voting machines closed so it’s not possible to check for security holes and other bugs, which lends credence to the various conspiracy mongers out there on the intarweb.
  • Large numbers of voters are put on and off the rolls by loosely supervised private contractors and other organizations often paid for by the parties, which leads to accusations of voter fraud (i.e., making up voters), voter exclusion (i.e., removing registration of eligible voters), etc.
  • Chicanery around electoral districts every ten years (or more often, as the case may be).
  • The use of partisan poll watchers to either intimidate (or unfairly challenge) voters or ensure voters vote the party line (by checking ballots for “validity”).

I’m sure there are others I’m missing but it really doesn’t matter, you get the point. These have more Republican fingerprints on them of late but Democrats shouldn’t be smug, you’ve been busted with your hands in the cookie jar before, too—the Republicans were just better at it than you were for a long time.

Part of the reason is relatively simple but often misunderstood: There’s always roaches in the walls and rat hair in your food. The closer you are too the cracks and crevices the more likely you are to see them. Well we’ve had close elections and close elections show you the cracks and crevices up close so no surprise there are roaches.

Stripping away all the civics class rah-rah, voting is a measuring system. It takes individuals’ subjective preferences, whatever they may be, and maps them into an aggregate choice of representative at various levels in a hierarchy. It cannot do this perfectly, as economist Kenneth Arrow showed in 1951, ever. In a nutshell, Arrow showed that five reasonable criteria (well, reasonable after a fair amount of thought) for decision making to be democratic are mutually incompatible. However, it’s not necessary to go to some of the esoteric extremes of Arrow’s mathematics to realize there’s a problem, and that problem is error.

Much like the price system—also a measuring system that takes your time, expertise and resources, my time, expertise and resources, and everybody else’s, even people you don’t know, and allows us to exchange them—it is ultimately subjective. Similarly the jury system is a measurement of the facts of a case when they are in doubt. Juries and voting are more similar than prices because in both much of the running of the system is in the hands of people who have stake in the “right” outcome happening and thus a lot of incentive to game it. (Of course there are regulations against price fixing for a reason.) Again, that matters in detail but not in the big picture. What is important is the fact that it is a measuring system and with measuring systems the important questions are:

  1. How much error are we willing to tolerate? Put another way, what level of resolution do we want from our measuring system? How many rat hairs are we willing to tolerate in a certain volume of peanut butter?
  2. How much does it cost to reduce error, realizing that because the universe is built on O(1/√n) convergence of averages, each additional digit of accuracy can be expected to cost increasingly more resources to attain? If rat hair free peanut butter costs, say, $100 per jar, who would buy it?

Let me give you an example to put this into perspective. The numbers quoted for DNA evidence are ridiculous. You hear numbers like “one in a million” or “one in a septuagenarian” or whatever. The actual number is unknown but probably somewhere around 1%… a reasonable guess for the probability of a human error in the chain of evidence. Human error, of course, dwarfs the chances of a real DNA match between people. If you mess up the evidence along the way, genetic science can’t help you. Sure Gil Grissom has a lower probability of error in chain of evidence, but even he’s not perfect and most CSIs are definitely not perfect, as OJ’s 1995 case showed the world. Assuming for convenience that human error and DNA sequencing machine error are independent,

Pr(one or both errors) = Pr(human error) + Pr(DNA sequencing error),

which effectively equals the probability of human error. Of course this doesn’t undermine DNA evidence, because in many cases the alternatives are worse, but it should make you a great deal more skeptical. In systems with substantial human involvement, we really can’t expect much better than about 1% give or take an order of magnitude… usually take an order of magnitude.

From actual data I saw of a study done in Illinois in 2002, optical scan with verify “kickback” (amusing use of that term in Illinois…), which is the best voting technology I know of, has an error rate of about… you guessed it, on order of 1%. So, unless we develop telepathic voting or are willing to spend a HUGE amount on the system, chances are good that 1% is about as good as it gets, or maybe one more digit, I would guess, but that’s just a guess. Maybe it can go smaller but no matter how you cut it, it’s not going to go smaller by all that much, which means that an election like 2000 in Florida—decided by a margin of about 0.01%—would be a nail biter for the foreseeable future. However, many people would, I believe, find the attainable level of error unacceptable, saying something like “It is a gross injustice if even one person is falsely denied the right to vote!” or “It is a gross injustice if even one person votes fraudulently votes when not entitled to!”

What’s ironic is that rather than taking effective, proven steps that would reduce the error level without breaking the bank (more on that below), we actually have many policies that increase the error level. The US is unusual in that voting is seen as a romanticized celebration of localism, civic virtue and so forth, and is handled largely by volunteers, many elderly, who are overseen by partisans. I don’t even need to leave the State of Florida to discuss this point:

  • If voting is so important it shouldn’t be left to the designer of the infamous Florida butterfly ballot, who violated a large number of principles of interface design because of well-intentioned ignorance.
  • Supervision of the system shouldn’t be left to elected officials who have multiple relationships.

Lest you get smug in other states, chances are good your electoral system sucks just as much but if things aren’t close it doesn’t matter. For instance, more chads were hanged in Illinois in 2000 than in Florida but because the vote in Illinois wasn’t close it didn’t matter, thus you didn’t hear about it. Then there’s the Land of Enchantment, where political pressure to bring cases of “voter fraud” (non-citizens voting or the good old standby of votes from bums “votes for smokes”, etc.) before an election got a good prosecutor kicked out of his job when he found no real evidence of it and declined to bring any charges.

It’s sad but many US elections simply don’t meet the standards that would satisfy electoral observers sent by, say, the Organization of American States (OAS).

To cut a long story short, I am arguing that we should turn the voting system itself over to a technocratic organization—let’s call it the National Electoral Institute, or NEI for short—built along the lines of the National Institutes of Standards and Technology, National Transportation Safety Board, Centers for Disease Control and Prevention or the Census Bureau. These three organizations are recognized around the world as the experts on their respective topics. If there’s an airplane crash somewhere in the world, chances are that an NTSB investigator will be helping the locals both because he or she knows a lot more about such things than the locals but also because NTSB is gathering more data about airplane crashes for when something happens in the US. Similarly, there basically isn’t a census in the world that doesn’t draw on the deep expertise of the Census Bureau. These organizations are non-partisan. (At least I hope they are and that there aren’t any Monica Goodlings or Kyle Sampsons floating around in them but Lord only knows after the last eight years.) In short: These are professionals who see it as their mission in life to make sure their jobs are done properly and according to spec.

One of the deep ironies is that the US government has pushed independent electoral institutes in other countries, e.g., Mexico’s IFE or the Electoral Commission of India, while letting our own electoral institutions languish. Yes we have the FEC but all it does is oversee campaign finance, and it’s far from clear that that’s even a good idea. But a true non-partisan institute would allow the United States, the world’s oldest extant democracy, to finally end Amateur Hour and assume a role as a shining example of how to do elections right, a role we should darn well have had years ago…but better late than never.

WordPress divider

Next Time: What the NEI would have as staff and function, barriers to making all this happen, and, of course, the downsides.