More research flaws in anti-election day registration study
One portion of the report issued by a prominent opponent of Election Day registration in Maine involves a chart titled “More Registered Voters than Voting-Age Citizens.” This has gotten relatively little attention, although it was mentioned by No on 1 advocate Rep. Jonathan McKane in a recent on-line chat sponsored by the Sun-Journal.
The chart (table 3) shows data for three elections in which the differences between the number of registered voters is above what they label “voting age citizens” — and the differences range from 4.974 to 57,495 people. The years included are 1992, 1996, 2004. If the large number of “extra” registered voters meant that there were that many votes cast in Maine while there were thousands fewer voting age citizens, that would be a problem, a real problem. It would mean fraud.
However, there are major research flaws.
One question one should always ask in designing a study is: What data do I need to help me understand a particular phenomena?
One wants data that are reliable and which are applicable to the phenomena under study.
There are, of course, many graduate courses and scholarly writings regarding this issue. This is because if one’s data are not reliable or applicable, then the data tells you — well, pretty much nothing.
One does not start by picking out data points but instead one tries to look at all the data in the period under study. If one selects data points from a long time period, there must be a method and rationale guiding the selection.
If one picks out data that “looks good” in terms of making one’s case, that is not research. It is cherrypicking. This might be done in the service of political advocacy or it could be an attempt to “prove” something for some other reason.
That said, what sort of data are included?
One sort of data are estimates of Maine people of voting age. These estimates come from the Census Bureau.
The other sort of data are numbers of Maine people registered to vote. These come from the Secretary of State.
Are these data reliable?
Reliability is at question for both.
The Census data are not the every ten year enumerations required by Article I of the U.S. Constitution, but estimates. The actual counts are often challenged for undercounts, with lawsuits filed by cities around the country. Moreover, as the Pew Research Center notes:
The population estimates, which update the census counts each year based mainly on government records, are easier for localities to challenge than the census counts. The number of jurisdictions with successful challenges to population estimates grew over the past decade, to more than three dozen in 2008.
And what of the the numbers of registrants? Registration numbers are from a time before Maine used a Central Voter Registry. As the Secretary of State’s website explains, “Historically, voter registration information in Maine has been maintained at the municipal level in more than 500 cities and towns throughout the State.” The Central Voter Registry follows a federal mandate to develop a database that “assigns a unique identifier to each legally registered voter in the State.”
In the time before the development of this system, it was much easier for any particular voter to be on the voter rolls of more than one place. Each voter did not have a “unique identifier,” so if one moved from Machias to East Machias, that person could be double-counted.
Truth be told, it is a bit hard to assess this situation because the data for all years are not presented.
But, taken together, the issues outlined above suggest that the Census, which is known for undercounting, likely underestimated the number of voting age individuals, while the registration totals, compiled at a time when multiple counts were far more likely, overestimated the number of people who were actual voters in the state. This very well may be the reason why the charts show more registered voters than voting age individuals in the state.
But there’s more. And this is critical.
Are these data applicable?
If the data are meant to demonstrate that there are currently more registered voters than voting age individuals, well, no. Recall, only three years were picked out — 1992, 1996, 2004.
Now, I remember 1992, but I have college students who were born before that. It’s quite awhile ago. And 2004? That’s seven years ago.
More to the point, these were years before the voter registry system was developed. This is key.
The sort of counts shown in the chart are from a bygone age.
And the figures, in any case, don’t demonstrate any fraud — either back then or now — since fraud is double-voting not double registration.
In fact, the chart doesn’t have anything to with election day registration at all — not in 1992, not in 1996, not in 2004, and not in 2011.
It doesn’t say what it’s purported to say.
The chart has data that’s neither reliable nor applicable to understanding Maine’s current voting system.