[EL] Comments on Doug Chapin's Blog Post

[Michael McDonald]

My apologies to Doug for being dragged in as a proxy to a debate that Curtis
wants to have with me, since he is obviously referencing my voting-eligible
population calculations.

I think Curtis has laid out the main issue correctly: Either you want a
turnout rate that is consistent with data available in 1860 or one that
provides the most accurate measure for modern elections. 

If you wish to compare current turnout to 1860, by all means use Curtis’s
numbers, if you like. These historical numbers have known inaccuracies. For
example, according to a 1997 U.S. General Accounting Office report, the
estimated net under-coverage of the decennial census was 5.8% of the total
population in 1940; 4.1% of in 1950, 3.1% in 1960, 2.7% in 1970, 1.2% in
1980, and 1.8% in 1990. That is to say, the older census were not counting
all people and thus historical turnout rates are inflated through a
deflation of a component of the denominator. So, while accurate historical
comparison is a laudable goal, we should acknowledge the deficiencies in
these historical numbers. Recognizing this, I see no reason to bind
ourselves to calculating turnout rates in a particular manner simply due to
the limited availability of 1860 data.

Prior to the 2001 publication of my voter turnout research in the American
Political Science Review (with Sam Popkin) Curtis did not calculate turnout
by citizen-voting-age population. He used voting-age population, which
showed turnout was declining when in fact the ineligible population was
increasing. Since then, he has shifted to using citizen voting-age
population, which I agree is a good step forward. Four years ago, Curtis
made a change to his historical southern numbers by removing
African-Americans. These are two major changes to his methodology that he
has made over the past decade. There is no reason why Curtis could not make
further refinements to his turnout estimates. I have frequently encouraged
him to do so in the most sincere manner possible. My bottom line is that we
need the most accurate numbers so that we can correctly determine what
effects turnout. This is particularly important, say, if one were examining
the change in turnout to a newly adopted election law.

If you wish to understand the determinants of turnout in modern elections,
you should use my voting-eligible population turnout rates which factor in
obvious and measurable populations such as ineligible felons. I welcome
constructive criticism since my goal is accuracy and I do make changes when
warranted, so invite Curtis to review my methodology more carefully. My
statistics are freely available to all on my website and I link to all my
canonical data sources. To briefly respond: 

- Naturalized citizens are included in my VEP estimates as they are based on
the ACS (as are Curtis's). 
- My most recent VEP calculations do include eligible overseas CVAP
estimates -- at the state level. 
- Recent movers are interesting to consider, but Curtis's 1860 interpolation
method does a poor job accounting for them, so he stands of shaky ground
criticizing me for an adjustment he does not make himself. The yearly July 1
Census population estimates better track the movement and growth of the
population by age among states from year to year throughout a decade, which
is why this is preferred data over a simple interpolation between censuses.
Still, I have been interested in making an adjustment for movers and to that
end I compiled the state laws regarding residency (an offshoot of this work
was published as "Portable Registration" in Political Behavior). I am still
contemplating how to address a mover adjustment. One thing to note here is
that due to the VRA requiring what are known as presidential ballots --
enabling people with the ability to vote for president regardless if they
moved after a registration deadline -- a mover adjustment would only apply
to midterm elections.
- Let's collect historical felony disfranchisement laws and make a
historical adjustment. Since the focus of my efforts are on modern elections
I have not prioritized that project. Maybe an enterprising law school or
graduate student would be interested...

My APSR publication completely changed the way in which an entire subfield
of political science and the media punditry world approached studying and
talking about voter turnout. Curtis was provided an opportunity to respond
to my APSR publication in that same journal, but he declined to do so. I was
offered an opportunity to respond in kind, too. I intended to show that even
using CVAP, turnout was not going down, which is what Curtis frequently
argued when he reported VAP turnout rates only. Certainly since then, Curtis
and I are in agreement that by either method of calculation for recent
elections we are near the top range of turnout for the past century. 

I applaud Curtis for the recognition of his book. I won't bore the list with
my accolades. I hope to announce several exciting new developments of my own
soon. 

Finally, unlike Curtis, who has already projected CVAP to 2020 by
extrapolating forward between the 2000 and 2010 censuses, I will use the
census current population estimates, ACS, and other data to continually
update throughout the coming decade to create the most precise VEP estimates
with the available data. As the decade progresses, our denominators are
going to become more divergent. For now, we are in somewhat agreement since
we both updated for the 2010 census, so I wish we could all just get along
for this short while.

============
Dr. Michael P. McDonald
Associate Professor, George Mason University
Non-Resident Senior Fellow, Brookings Institution

                             Mailing address:
(o) 703-993-4191             George Mason University
(f) 703-993-1399             Dept. of Public and International Affairs
mmcdon at gmu.edu               4400 University Drive - 3F4
http://elections.gmu.edu     Fairfax, VA 22030-4444

From: law-election-bounces at department-lists.uci.edu
[mailto:law-election-bounces at department-lists.uci.edu] On Behalf Of Curtis
Gans
Sent: Thursday, January 05, 2012 10:43 AM
To: law-election at uci.edu
Subject: [EL] Comments on Doug Chapin's Blog Post

There is an alternative to what Chapin has chosen to put in his blog as the
choices for numerators and denominators for assessing turnout. I use a
denominator that was pioneered by Walter Dean Burnham -- which is neither
VAP nor VEP. It is citizen-eligible population -- age-eligible minus
non-citizens interpolated between Censuses. I believe it is a better
denominator for four basic reasons: 1. It removes the largest distortion in
the age-eligible (VAP) figures. 2. It is the only denominator for which
there is comparable data going bask to 1860. 3. VEP includes only some of
the elements that affect the denominator (non-citizens and felons). It does
not include Americans living outside of the United States who are not
included as age-eligible but can vote, people who are naturalized in the
year of the election, for some elections overcounts and undercounts and
movers during the year of the election. 4. Beyond non-citizens, there is
only limited historical data for many of the factors  -- one can find data
on convicted felons back to the 19teens, but there is no compilation of law
as to which of those felons or ex-felons could by law vote. Undercounts and
overcounts have only been assayed since 1942. It is possible to accurately
count those citizen living outside the United States who are in the military
and in government positions and allocate them to states. It is virtually
impossible for the others to be so estimated and allocated. Naturalization
figures and movers are only available after the election year. For all of
these reasons, I believe the figure I use and many others -- age-eligible
minus non-citizens is a better and more historically comparable measure. It
is the denominator used in my book -- Voter Turnout 1788-2009, published
late 2010 by CQ Press, which was juried and which was selected as one of the
23 "outstanding reference books of 2010 by the American Library Association.

One last point, total ballots cast would, as Chapin suggests, be a more
accurate numerator for true turnout performance. It is a figure I collect.
But it is not available historically for many states beyond the past few
elections. Which is why I and most others use presidential vote for
presidential years and highest turnout race (including aggregate U.S. House
vote in states with no statewide top of the ticket races in any given year)
for mid-terms.