MyDD :: NJ GOP lies, NYTimes Bites

NJ GOP lies, NYTimes Bites

by Kombiz Lavasany, Sat Sep 17, 2005 at 08:27:06 PM EST

The New Jersey Republican party learned how to use a simple database this last week, and the New York Times decided to let the GOP write the story without an attempt to verify the GOP's press release. The story is ominous enough, thousands of dead voters in NJ voted in the last election, and thousands of people living in multiple counties voted in more than one county. Almost 7,000 voters voted in NJ voted in other states. The whole article has an ominous ring to the story, except for the fact the New Jersey GOP says it's unwilling to accuse anyone of any wrong doing, or say whether the voters are registered Democrats or Republicans.

This is how they ran their query:

Republicans analyzed voter records from the state's 21 counties, and matched those voters whose first and last names and dates of birth matched in more than one county. Of that group, nearly 4,400 of those matches voted in two counties. Middlesex County had, by far, the greatest number of potential duplicate voters, with 2,299, according to the analysis.

Why won't they accuse anyone? Election expert Michael McDonald explains in an email to the election law listserv.

I would have thought we would have learned from Florida's experience of matching a list of felon names against their voter file that one should be cautious when matching names and birth dates.

I happen to have at my disposal for research purposes voter registration files with 2004 vote history for several states, including Ohio, but unfortunately not New Jersey. I decided to do the same match as the New Jersey Republicans for the state of Ohio. I found 9,108 instances where voters had the same first name, last name, and birth date. However, there were 6,498 matches on `1/1/1800', which some Ohio counties use to identify missing birth dates.

...Finally, I assume that New Jersey like most states records phone number on their voter registration files. It would not have been difficult for the Republicans to call a few people to verify that they were truly alive or different people. Such hard evidence of fraud would have left no doubt in anyone's mind, and that the New Jersey Republicans did not take this step or at least present findings from this step makes me highly suspicious of their claims.

There you have it, the NJ GOP decided they would get some press by floating the idea that there was massive voter fraud in NJ, and the New York Times ate the entire press release up without telling you the entire process was merely an experiment in running a query in a database. The only interesting factoid from the story is now we know the answer to the burning question: In a state of 8 million people I wonder how many other people have the same first name, last name and birthday?

Tags: Republicans (all tags)

Permalink :: 34 Comments :: Post a Comment

The classic "Birthday Problem" (none / 0)

There is a classic math problem called the "Birthday Problem", which asks, how many kids to you need in a class before 2 of them are likely to have the same birthday. The answer is that at 23 kids, you have a 50% chance of having 2 of them with the same birthday. (See

http://www.pen.k12.va.us/Div/Winchester/jhhs/math/lessons/stat/birthday.html

for a reference.

Well, I used Google to look up how many people named "James Smith" are in NJ. There were 217. Since some are unlisted, and not all are registered, I assumed 140 were registered. If you assume they have birthdays over 50 years (18-68) there is a 46% chance that two of them will have the same birthday.

For a name with 100 voters, it's 24% chance of a common birthday. For 20 voters, 1% chance. I don't know, 100 first names, 500 last names, 50,000 combinations at 1% chance of a duplicate gives 500 duplicate first/last/bithdates?

I'm not a mathematician, and I don't know "name demographics", but I think its not unreasonable to expect 1000-2000 duplicates in the state?

Matt - DemConWatch

by msn1 on Sat Sep 17, 2005 at 09:35:25 PM EST

Re: The classic "Birthday Problem" (none / 0)

Also, people move and the registrar of voters doesn't always do good cleanup. Shocking, that. If, for example, someone moves from NJ to PA for a year, registers & votes there, then returns to a different county in NJ and reregisters, then what?

by paperwight on Sat Sep 17, 2005 at 10:18:45 PM EST
[ Parent ]

Re: The classic "Birthday Problem" (none / 0)

The numbers that we are discussing here are not duplicate registrations, which would yield a number far higher than 7,000. The numbers that are being discussed are those where people who were registered in more than one locale voted in more than one locale. Your hypothetical voter would not be counted in the total.

by Mr Moderate on Sat Sep 17, 2005 at 10:26:49 PM EST
[ Parent ]

Re: The classic "Birthday Problem" (none / 0)

Nevermind. Disregard the above.

by Mr Moderate on Sat Sep 17, 2005 at 10:34:20 PM EST
[ Parent ]

Re: The classic "Birthday Problem" (none / 0)

In one comment you say the number can't be greater than 4, and in another you say it's greater than 7,000.

by Kombiz Lavasany on Sat Sep 17, 2005 at 10:46:31 PM EST
[ Parent ]

Re: The classic "Birthday Problem" (none / 0)

Huh?

"In one comment you say the number can't be greater than 4..."

That number being the number of people who will legitimately have the same birthday, last name, and first name for a state of five million registered voters.

...and in another you say it's greater than 7,000."

That is the number of people here who, according the NJ GOP, have the same birthday, last name, and first name, but voted in the last election.

7,000 is much larger than four, I agree. WHICH IS THE PROBLEM. If the number of people who voted sharing all three traits is statistically larger than four, then fraud (or some other problem) exists.

by Mr Moderate on Sat Sep 17, 2005 at 11:02:25 PM EST
[ Parent ]

The problem is your math. (none / 0)

See below.

by Drew on Sat Sep 17, 2005 at 11:10:38 PM EST
[ Parent ]

Re: The classic "Birthday Problem" (none / 0)

I did this problem in college and the answer was if you had 18 people together, the odds were better than 50% that two of them would have the same birthday.
In the example cited above, they use a 366 day year, we used 365. If you want to be a stickler you'd use 365.25.

by Ryan on Sun Sep 18, 2005 at 03:51:38 AM EST
[ Parent ]

Mathematical based analysis (none / 0)

When I registered to vote in New Jersey, I was not asked for a phone number or asked what party I wanted to register under.

Today, I know 'phone number' is now a field on the form, but it is optional. And I know that a lot of people don't fill it out, because it's just an invitation to get non-stop telemarketing calls from late October through Election Day.

It's not as simple as to just call everyone (and besides, who is going to admit voter fraud to some guy calling on the phone?). But, if you just do some simple math, you'll see that it's probably not necessary to call:

"The only interesting factoid from the story is now we know the answer to the burning question: In a state of 8 million people I wonder how many other people have the same first name, last name and birthday?"

Are you insinuating that there are absolutely no fraudulent double registrations in the state of New Jersey? That's ridiculous. Washington, Ohio, Missouri, and Florida should have taught you that voter fraud is rampant on both sides, and it's laughable to think that New Jersey would be any different. I don't get why you'd argue otherwise -- exactly who are you trying to defend in saying there isn't any fraud?

As a quick exercise:
The most popular male first name is James, which 3.318% of the US shares; 2.629% of females are named Mary.

Smith is the most popular last name, which clocks in at just 1.006%.

And, of course, on average, 0.27% of people will share the same birthday (1/365.25).

In New Jersey, 4,797,345 people are registered to vote as of the 2005 Primary.

Statistically, putting all that together, as a best case scenario designed to get the highest possible number:

0.03318 * 0.01006 * 0.002738 * 4,797,345 =

... (drum roll please)

...

... a fraction over four.

On average, in a state with the population of New Jersey, four people (or 4.384, to be exact to four places) will have the same first name, the same last name, and the same birthday.

by Mr Moderate on Sat Sep 17, 2005 at 10:21:03 PM EST

Re: Mathematical based analysis (none / 0)

Thanks Mr. Moderate now read the post over again. The NJ GOP did a study (read ran a query in a databse) and found out that so many people had the same name and birth date in different counties, or the same name as someone who died, and then said these are potential double voters - look vote fraud in NJ. Except they don't accuse anyone of double voting, or can't say anyone double voted because they know their "experiment" was crap, and meant only to muddy the waters. McDonald ran the same experiment in Ohio, and other states and came up with a few thousand. If you don't think the NJ GOP doesn't have those people's phone numbers you're are mistaken or being obtuse (they get them from marketing lists).

by Kombiz Lavasany on Sat Sep 17, 2005 at 10:29:20 PM EST
[ Parent ]

Re: Mathematical based analysis (none / 0)

"The NJ GOP did a study (read ran a query in a databse) and found out that so many people had the same name and birth date in different counties, or the same name as someone who died, and then said these are potential double voters - look vote fraud in NJ."

Of course they are potential double voters. The opportunity for fraud is wide open there. Are you saying it's not?

"McDonald ran the same experiment in Ohio, and other states and came up with a few thousand."

Because, as you said, Ohio uses a false birthdate for folks who do not enter their birthdate. Besides, I don't understand the arguement here -- the potential for vote fraud exists in New Jersey...but that's okay because it exists in Ohio too?

"Except they don't accuse anyone of double voting, or can't say anyone double voted because they know their "experiment" was crap, and meant only to muddy the waters."

What's crap? What water are they muddying? I don't see anyone disproving this allegation here. These are serious allegations that deserve more than a simple dismissal. And it CERTAINLY can't be explained away as simply as you've tried, saying that seven thousand people in New Jersey have the same first name, last name, and birthday.

"If you don't think the NJ GOP doesn't have those people's phone numbers you're are mistaken or being obtuse (they get them from marketing lists)."

I've worked more voter ID than any human on earth should ever be forced to suffer through, and I've worked as a field manager. I know how god awful and incomplete those phone lists are.

Of course, they could have called some people, but they certainly don't have to. And what here exactly is the NJ GOP lying about?

by Mr Moderate on Sat Sep 17, 2005 at 10:53:05 PM EST
[ Parent ]

Re: Mathematical based analysis (none / 0)

Mr. Moderate,

you are presuming that all the variables are independent. To dummy test this theory of independence, compare the fraction of people named "John Smith" with those named "John" and those named "Smith". I'm sure you'll find the names correlate.

If you want to do a statistical argument, you should make sure it's sound first.

by RickD on Sun Sep 18, 2005 at 05:38:25 AM EST
[ Parent ]

Your math is wrong. (none / 0)

What you actually calculated is the likelihood that, if you drew a person from the New Jersey voter file, that they would have a specific first and last name and that they would have a specific birthday.

That's much lower than the number of voters with the same name and birthday.

by Drew on Sat Sep 17, 2005 at 10:58:49 PM EST
[ Parent ]

Re: Your math is wrong. (none / 0)

No, that's incorrect.

To have two people match here, you must first draw a person who has a specific first name, last name, and birthday.

Now, if a second person pops up with the same first name, last name, and birthday, there are two possibilities:

(1) Both are the same person. We know for a fact that this does happen, and since there's no mechanism to prevent it, how can you discount it?

(2) They are different people who just happen to share all three traits. The odds of this happening are STILL 0.00009139%, or one in 1.1 million.

by Mr Moderate on Sat Sep 17, 2005 at 11:12:54 PM EST
[ Parent ]

Re: Your math is wrong. (none / 0)

Listen I'm going to do this very slowly for you as a list. It could be me, I could have a problem articulating something very simple, but I suspect it's you since you like defending the NJ GOP here.

NJ is a state in the Northeast with 8 million people
NJ GOP learns how to use a relationshal database
NJ GOP intern discovers that so many people in NJ share the same name and the same birthday
NJ GOP intern runs their voter file against voter files in other states and discovers more people who have the same name and birthdays voted in the last election (thank the RNC for this)
NJ GOP does a press release about fraud in NJ
NJ GOP tries to capitilize by doing saying there are so many thousand potential fraud cases in NJ.. even though they are alleging fraud they won't back it up - That's the GOP lying
NJ GOP refuses to do follow up because they know that those people are real voters in NJ.
NYT runs their press release as a story without noting the fact the whole premise of their press release is unscientific and univerified.
Election Law professor notes that this is to be expected, and runs his own test with names and birthdays in the states where he has access to a voter file
Quick experiment by professor who knows elections/election law/voter registration database reveals the underlying problem with the NJ GOP press release

by Kombiz Lavasany on Sat Sep 17, 2005 at 11:49:31 PM EST
[ Parent ]

No, I'm right. (none / 0)

And you don't know what the fuck you're talking about.

Is it unlikely that there are two James Smiths who were born on January 1st? Yes. Two James Smiths who were born on January 2nd? Yes. Two James Smiths who were born on January 3rd? Yes. And so on through every day of the year and every name in the state. You know what? Eventually, I'll hit two people with the same birth date and the same name. And I may very well hit it a few thousand times out of a few million.

I sure as hell will hit it more than four fucking times.

by Drew on Sat Sep 17, 2005 at 11:52:51 PM EST
[ Parent ]

Re: No, I'm right. (none / 0)

Drew is right. The key is that any day can match, not just a specific day. That's the whole point behind the birthday problem, you only need 23 kids to have a better than average chance of 2 kids having the same birthday. And Moderate's analysis also assumed the same birthday, without regard to the year. I'm sure the year was also take into account, so so he would have to divide by another 50 years or so, making his number essentially 0, which I think is clearly wrong. Check the math in the article - it applies here, and the number is probably in the 1000-2000 range.

Matt - DemConWatch

by msn1 on Sun Sep 18, 2005 at 12:26:51 AM EST
[ Parent ]

Re: Your math is wrong. (none / 0)

Mr Mod

Your math is indeed incorrect. By your own admission, you're calculating the likelihood that a given voter will have another voter match their first name, last name and birthday. This is very different from calculating the chance that any two random voters will match all three criteria.

Consider the classic birthday problem described above. If you have a group of 23 people, there's about a 50% chance that two of them will share the same birthday. The formula you used, however, would calculate the odds of someone sharing, say, your birthday. Obviously, 23 is much too small a number for the odds to be 50%.

by gorillagogo on Sun Sep 18, 2005 at 12:25:30 AM EST
[ Parent ]

Re: Your math is wrong. (none / 0)

Mr Moderate: gorillagogo has the basic idea here. You're calculating the odds of the match between a specific pair of voters. Now you have to sum over all possible pairs of voters.

With 8 million voters, there's 32 trillion such pairs.

To get 7000 matches amongst that number of pairs, you just need a 1-in-4.5 billion random chance of a match for any given pair.

Birthdates and names should be independent of one another. Say we have 22,500 distinct birthdates (approximately the number of birthdates for everyone over 18 but under 80); the odds against a name match would have to be 200,000 to 1. So if the typical first-name, last-name combination is repeated 40 times amongst the NJ voter records, the 7000 matches are to be expected.

by RT on Sun Sep 18, 2005 at 06:36:10 AM EST
[ Parent ]

Re: Your math is wrong. (none / 0)

Mr. Moderate,

you should study the math of database queries. You're making a fairly elementary mistake. You are essentially testing the probability that the first two people you look at match. (Except that you make a dubious distinction regarding independence of "first name" as "last name".) That's a much lower probability than "two people match". The basic problem is that each new person added to the database would be counted as a match if he matched any of the people listed before him. So the probability of avoiding matches decreases as the database grows.

Of course, you're just doing what the NJGOP intern did: trying to allege wrongdoing without any actual evidence of it. Find two registrations that match the same voter, who voted twice. If there are so many matches in the database query, this should not be hard to do, right?

by RickD on Sun Sep 18, 2005 at 05:48:20 AM EST
[ Parent ]

Re: Mathematical based analysis (3.00 / 1)

Congrats, you have just calculated that there will likely be about 4 James Smiths with the same birthday on the registration rolls in NJ. Now you need to calculate how many Jerry Jones will share a birthday, how many Mable Jenkins share a birthday, How many John Schmerzs share a birthday, etc., etc.

by Dave Howard on Sun Sep 18, 2005 at 12:24:14 AM EST
[ Parent ]

Re: Mathematical based analysis (none / 0)

On average, in a state with the population of New Jersey, four people (or 4.384, to be exact to four places) will have the same first name, the same last name, and the same birthday.

You forgot add "and whose first name is James Or Mary and whose last name is Smith" That might have more than a few implications for your "analysis"

by boswell on Sun Sep 18, 2005 at 12:58:44 PM EST
[ Parent ]

NJ Politics--truly many bloodsuckers (none / 0)

One assumes here that the NJ GOP is actually competent enough to figure this all out. That I doubt, being a former NJ GOPer--right now, the party is bereft of much in the way of leadership or skill.

NJ Dems aren't the best or the most competent; NJ government is probably the 2nd most corrupt in the country.

by billcoop4 on Sat Sep 17, 2005 at 11:02:26 PM EST

Re: NJ Politics--truly many bloodsuckers (none / 0)

I wonder if part of this was to muddy the waters on the election fraud issue.

I also worry about the Seeing The Forest Rule, when you see a Republican accusing others of something, it means they are probably doing that thing

by Alice Marshall on Sat Sep 17, 2005 at 11:20:01 PM EST
[ Parent ]

Re: NJ Politics--truly many bloodsuckers (none / 0)

In New Jersey, assume all parties are doing it.

by billcoop4 on Sat Sep 17, 2005 at 11:53:34 PM EST
[ Parent ]

The Story is a Support Action (3.00 / 1)

I've seen some sample Republican ads for NJ in the past few days: message is "Time to end the Corruption." They are going to try to paint the Democrats as corrupt. An honest defense would be "of course we are, just like the Republicans, but at least we're not nuts!"

by Jim P on Sun Sep 18, 2005 at 01:15:21 AM EST

How NJ Dems should counter (3.00 / 1)

The NJ GOP's claim is ridiculous, though perhaps they weren't clued enough in math to realize it. But explaining the math is not an effective way of debunking this story once it gets out, which it seems to have thanks to the NY Times. So if I were the NJ Democratic Party, here's what I'd do:

Search the voter file and find some voters in different counties who have the same first and last names and birth dates. Make a list of Republicans who share these characteristics. Limit the list only to those who voted in the 2004 presidential election.

Contact a few of these people, and find a pair, or two, who are verifiably different people. Tell them about each other's existence, show them the NY Times story, get them in touch with each other, get their reaction.

Shop the story to the NY Times. Get them to write letters to the editor, or write an Op-Ed about it, or something.

by cos on Sun Sep 18, 2005 at 02:21:35 AM EST

are they checking birth-days or birth-DATES? (none / 0)

Much of what is written above is pointless if we are talking about matching birthdays (February 5), while the analysts are comparing birth DATES (February 5, 1937).

While not impossible, it is far, far, less likely for two people in the same state to have an identical name and birth date.

by mathchallenged on Sun Sep 18, 2005 at 03:27:35 AM EST

Re: are they checking birth-days or birth-DATES? (none / 0)

I wouldn't say it's "far, far less likely". If you restrict your attention to the voting age population, the probability of being the same age as another voter is, what, 1/40? Worst case 1/80? And that's for people at the extreme end of the spectrum. You're likely to get more matches for younger people.

Reverse the question: how many people named "John Smith" are born in a given year?

by RickD on Sun Sep 18, 2005 at 05:52:02 AM EST
[ Parent ]

Not Necessairly (none / 0)

The SSA keeps up with the most popular 1st names in any given year at least as far back as when I was born. The point being that naming children seems to be a bit faddish. So in any given birth year you are going to have a lot of common 1st names. Tack them on to a common surname such as Smith or Jones and it doesn't seem far fetched to me. Also potential fraud is bullshit. Potential is not actual. Prove that both of the "identical" named parties actually cast 2 ballots and were actually the same person.

"Once in a while you get shown the light In the strangest of places if you look at it right"

by molly bloom on Sun Sep 18, 2005 at 10:41:58 AM EST
[ Parent ]

Why complain? (none / 0)

I don't understand your objection to this kind of analysis; it seems to me the majority of people with the mobility and resources to move frequently would be disproportionately Republican anyway. A national ID card would cure a lot of these problems.

by wayneseib on Sun Sep 18, 2005 at 08:45:02 AM EST

Don't know about other states... (none / 0)

But, at least in Jersey, it's the opposite. Democratic voters tend to be more mobile, and tend to follow better jobs around the state. There are many transitory voters in the larger cities. You don't need means to move around--we're also talking moves between apartment buildings, large cities, etc.

I see what the GOP is saying, though I suspect that they are (gasp) exaggerating for effect and to score political points. Since the New Jersey voting population has a lot of low-to-middle income families, as well as a lot of young professionals, there is significant migration into and out of the state. Interstate cooperation being what it is, it's highly likely that those duplicates between states could remain on the voting rolls for a significant period of time, just because when people move, they don't always think to alert their old county, and they don't always remember what address they were registered from, or if they even were registered to vote (maybe they signed a form and forgot about it).

It's a good argument for federalizing the voting rolls, and creating uniform standards and forcing states to cooperate with each other in order to prevent this sort of duplicate registration which, while not malicious on the part of the voter or inherently wrong, is open to abuse by people without scruples. (Of course, this isn't why the NJGOP is bringing the issue up, and for the most part, I don't think their leadership actually cares about the issue at hand, but just about the political points that can be scored off their findings.)

by sucopsucoh on Sun Sep 18, 2005 at 12:24:46 PM EST
[ Parent ]

sad misunderstanding of math (none / 0)

This shouldn't even need to be discussed to point out that it's utterly laughable -- it's clear to anyone who even gives it half a thought that in a state of several million people you're going to find some people born on the same day with the same name.

How many baby Smiths or Martinezes or Johnsons are born in the entire state of New Jersey every day? How many of those babies with the same last name just happen to be given the same first name too?

When you consider the odds, it doesn't seem too unlikely an occurrence given the size of the population. Now consider: To reach 8000 over fifty years, you'd just have to have babies get the same name by chance once every couple of days.

This is about as transparent a bit of fear-mongering as it gets. The NJ GOP is pushing the standard GOP line that those damn liberals are stealing the elections. (Well, "potentially" so. If you squint at the numbers really really hard, until you can't actually see them any more.) The point? Probably to justify to themselves all the voter-suppression work they do. After all, in their twisted little minds they can justify playing dirty and undermining democracy if they believe the other side is playing dirty too.

BONUS QUESTION: Why didn't the NJ GOP didn't say how the sample was broken down by party?

BONUS ANSWER: Because the set of people with duplicate names+birthdates probably actually included a larger proportion of Republicans than the state overall. The fashion among black families in the past twenty years or so has been for creative names that are more likely to be unique, which would lead to an underrepresentation of blacks in the sample.

by drewthaler on Mon Sep 19, 2005 at 09:53:03 AM EST

go scarlet knights! (none / 0)

please don't overlook the fact the middlesex county is home to rutgers, i.e., thousands of students who are in middlesex county but make their home in the other nj counties. now add dozens of well-meaning volunteers who hang around campus registering students to vote and a few thousand absentee ballots to the mix; the resulting statistics can easily be used to portray an ominous situation... i wouldn't be surprised if i'm on the books in both camden county and middlesex county.

by big in japan on Mon Sep 19, 2005 at 02:00:15 PM EST

You are not logged in.

In order to post a comment, you must be logged in. If you have a member account, please log in to comment.

If not, you can make an account right here. It's quick and free.