[EL] Is Minority Registration Really Declining?

[Doug Hess]

Some thoughts on the CPS and Washington Post debate, although most of these
do not address the accuracy of the Post article (which initially explained
registration purging incorrectly, too):

1. My first thought on the Washing Post article was that they were ignoring
sample size problems of subgroups within states. True, the CPS is huge
(roughly 45,000 households nationally) and it is designed to have a
representative sample of state households (although see below for
representativeness). Nonetheless, to look at minority registration rates by
state, you are making the following reductions in the data: remove all
non-citizens, remove all non-adults, cut the sample into 50 states plus DC,
cut the sample into race/ethnic categories. When you look at a sample that
has been reduced by that much you may get some large standard errors of
difference, which is what you need to make year to year comparisons.
Unfortunately, another issue is that the way the CPS staff want you to
calculate standard errors is a bit complicated (I don't think many people
follow it). So...I haven't done the math, but my first thought about these
year to year variations by state by ethnicity/race was that they were
assuming too much.

2. The article mentions residential mobility as a possible source for a
perceived decline in registration. Of course, people in the US move a lot
(the data on residential mobility really are amazing) and length of time at
residence is one of the strongest predictors of voting. However,
residential mobility tends to decrease when the economy is weak not
increase. People often move for work. Thus, when unemployment is high,
moving declines (a bit). Still, if home foreclosure problems are large
enough (I just don't know) it would be an issue.

3. The non-response problem in the CPS does need additional research as
Michael points out. However, perhaps imputation would be a better
alternative to deletion? It's a sticky problem. I'd be excited to see any
research on this, including information on who does not respond and why.
Other supplements to the CPS have a pretty good response rate, so why are
there so many non-responses to the November Supplement (especially
considering how short it is)?

I wonder if there is evidence regarding why people don't respond that
supports the long history of treating non-responses as non-registered and
non-voters. Treating them as non-voters does seem to provide data that
lines up, nationally at least, with exit polls on turnout for race (if I
recall correctly...).

4. Another rarely discussed issue is that the reporting in the CPS is by
proxy. That is to say, one respondent in the household responds for
everybody else (i.e., the household respondent will be asked about the
voting and registration status of their spouse, adult children, etc. in the
household). I know one paper looked into this, but this aspect might need
more examination also.

5. Keep in mind that the CPS does not include populations that are in
"institutional" housing (military barracks, nursing homes, college
dormitories, shelters, etc.) and does not include the homeless. So, when
matching what it shows to the number of ballots actually cast, etc., there
is that issue.

Anyway....more could be said. The CPS remains a great resource, but needs
to be used carefully. And a revamping of the CPS November Supplement is
overdue.

I wonder if some large polling firms ask about voting in November? I don't
mean campaign polling, but the firms that ask 100,000 people each year
about what mustard they use, how often they eat out or go see a movie, etc.
These firms already ask a bunch of other questions about health, etc. in
addition to their consumer questions. Perhaps voter turnout would interest
them? It could get a lot of play.

-Douglas R. Hess, PhD
douglasrhess at gmail.com


---------- Forwarded message ----------
From: Michael McDonald <mmcdon at gmu.edu>
To: law-election at uci.edu
Cc:
Date: Mon, 07 May 2012 17:48:15 -0400
Subject: Re: [EL] Is Minority Registration Really Declining?
Here's a link to my HuffPo critique of the Washington Post story asserting
minority registration rates are declining:

http://www.huffingtonpost.com/michael-p-mcdonald/is-minority-voter-registr_b
_1497813.html<http://www.huffingtonpost.com/michael-p-mcdonald/is-minority-voter-registr_b_1497813.html>

They are not, if anything Hispanics are slightly up from 2006 and Blacks are
significantly up. My issue is with the CPS -- the source for the story -- is
that the CPS includes persons who do not answer the voting registration
questions as not being voted or registered. These non-responses should more
appropriately be treated as missing data and excluded from the rate
calculations. This insight resolves for me other puzzling issues that I've
recently had with the CPS, such as apparently declining CPS turnout rates
when they are most certainly increasing.

Those familiar with my voter turnout work will appreciate this:
non-respondents to the CPS voting and registration questions are increasing,
CPS turnout and registration rates are not declining. But, as an add-on,
unfortunately the corrected CPS turnout rates demonstrate significant vote
over-report bias, on par with the ANES.

Btw, Chris Achen and Jon Krosnik independently turned me on to the CPS
non-response issue. I am co-author with Jon and colleagues on a manuscript
that delves more deeply into over-report bias on the CPS and other surveys.

============
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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://webshare.law.ucla.edu/Listservs/law-election/attachments/20120508/0a8f019f/attachment.html>