[EL] “Yes, ‘Citizens United’ gives Republicans an electoral edge. Here’s proof.”

Sat Apr 9 10:50:45 PDT 2016

Here is my critique of “The Business of American Democracy: Citizens
United, Independent Spending and Elections.”  I read the July 2014 version
up on SSRN, hopefully there isn’t a newer version.

GENERAL MODELING

They define “a race” as “all primary, general and runoff elections
associated with its office, district, state and year.”  This is their unit
of analysis, and this choice of unit of analysis is highly problematic.  At
first, I had a hard time believing that this wasn’t a typo, and it is only
stated once in the paper.  However, a collection of other evidence as the
paper progresses indicates that this is not a typo and I am reading this
correctly.  For example, on page 13 they state that the sample had an
average of 2.259 candidates.  Since these are in one seat races (confirmed
below), there are a few possibilities.  The first is that there are a lot
of non-major party candidates.  Even if there was one Democrat and one
Republican in each election (which there obviously aren’t), that would
imply there are .259 non-major party candidates per election sounds too
high to me.  I know there aren’t that many.  Next, in Table 2 they report
that there are more than one Republicans and more than one Democrats in the
average election.  Perhaps this is caused by the top-two primary?  In CA
and WA, you can get two dems or two repubs in an smd, but that wouldn’t
bring the average up enough to offset uncontested elections.  Neither of
these states had bans, so they would both be in the middle two columns of
Table 2, and the right most two columns still have more than one on
average.  They really do lump primary candidates in with general election
candidates.  Why this is bad will be returned to when discussing Tables 3,
4 and 5, which each have a different set of dependent variables.

Table 3

They failed to mention in their op ed or in the abstract that they didn’t
find effects for state senates.  In the paper, they don’t discuss at all
why they’d find an effect in state house elections, but not state senate
elections, for examining the impact on Republican success.  If anything,
state senate elections would be more amenable to influence by independent
spending.  My general observation is that the larger and more powerful the
office, the more subject it is to idiosyncratic factors.  For example,
prediction models have an easier time predicting U.S. House elections than
U.S. Senate or gubernatorial elections.  By that logic, state senate
elections would receive a little more attention than state house elections,
and voters would be more swayed by non-partisan factors, such as the
individual attributes of the candidate.  That said, I think this difference
would be trivial when evaluating which party wins an election.  But if
there was a difference, it would go in the other direction.

Their modeling strategy in Table 3 is problematic.  They examine the
probability of a Republican winning even when there are no Democratic
opponents.  But it is a certainty under that condition that the Republican
will win.  Y is built into one of the Xs, making this a tautology.

They deal with redistricting by restricting the sample to 2002-2010,
instead of 2000-2012.  They acknowledge that redistricting will be a
problem for the district fixed effects.  But merely restricting the sample
is going to miss some redistricting.  Looking at the State Legislative
Election Returns database, the following state houses redistricted in 2004;
AZ, FL (minor), GA, ME, MA (minor), MT, NH, NC, ND.  AZ and ND aren’t in
sample, so no problem for them.  That leaves four SMD states plus NH, that
has some SMDs.  Only 12 seats in NH are SMDs that redistricted, so no
problem for them.  So that leaves GA, ME, MT and NC.  NC and MT had
corporate bans, the other two didn’t.  This means that 4 out of 46 states
are totally messed up in their sample (8.7%).  The same states are affected
for state senates, except add 18 messed up seats in AZ, because the AZ
state senate has SMDs and 18 out of 30 seats were re-redistricted.

In table 3, they find that the more Republicans run in “an election” (i.e.,
a more accurate way to say this is “the more Republicans contest the
primary before the general election”) the more likely a Republican is to
win the general election.  It is obvious why this would be the case.
Republicans are more likely to contest a primary where Republicans do
better in the general.  Same for strongly Democratic districts.  What is
not obvious is the consequence including this variable has on the impact of
“IE ban x Post-CU” which has a SE of .020, and a coefficient of .041,
meaning it’s right on the edge of statistical significance.  It also means
that for another reason, a tautology is being built into the model.  Actors
a few months before the general will predict the outcome of the general and
contest the primaries accordingly.  Again, this means that a large aspect
of Y will be built into an X.

Table 4

Tables 4 and 5 don’t restrict the sample to 2002 to 2010, unlike in Table
3.  So they don’t even inadequately deal with redistricting there, they
simply don’t deal with it at all.

They don’t take term limits into account, which is highly problematic for
their analysis of incumbent retirements.  An incumbent who was going to be
forced into retirement by term limits should not be included in the
analysis of retirements.  No X can affect such an incumbent’s decision.

Their model of incumbent retirement has an R2 of .01, which is extremely
low, and has no control variables.  District size, the professionalism of
the legislature, how safe the district is, and whether there was
redistricting or not would be major determinants.

Why would being more likely to be the beneficiary of independent spending
cause republicans to be less likely to retire in districts that are safe
anyway?  There should be an interaction between post-Citizens and how safe
a district is.  No one would argue that less than 80% of districts are
totally one-sided, so this is a major issue.

Table 5

Combining primaries and general elections renders the results of Table 5
meaningless.  So what if Democrats are less likely to contest primaries?
The setup of the paper is about general elections and the impact of
corporate independent spending on the relative electoral success of the two
major parties.  The health of internal party democracy is not at issue.

Table 5 also appears to lump in non-major party candidates.  If you add the
average number of Democrats and Republicans in Table 2, it doesn’t reach
the average number of candidates, so .176 candidates an election are
non-major party candidates.  It is not very significant if an independent
that obtained 5% of the vote ran or not.  This is fundamentally different
than if a Democrat failed to run against a Republican.

MINOR COMMENTS

They divide the population of a state by the number of unique districts to
get a measure of district population.  They need this to measure per capita
expenditures in districts.  This causes them to misestimated district per
capita spending in states with different numbers of seats in different
districts (state houses; VT, NH, MD, and WV.  State senates; NC (2000 only)
and VT) and they do not take floterial districts into account (NH house
only).

They don’t take union density in a state into account.  If unions are
strong in a state, the impact of removing citizen’s united bans in a state
on Democratic fortunes would be greater.

They only use a dichotomous variable to track whether people won or not,
not a % of the vote variable.  This isn’t problematic.

They use ols for everything, even though Y is dichotomous.  Not too
problematic.

They exclude races with more than one winner, meaning they excluded
ffa-mmds, which is not problematic.  However, they say that they have 47
states in their sample for house races.  That doesn’t sound right.  First,
they already excluded NE, so that’s 49.  States with only ffa-mmds for
state house; az, sd, nd, which makes 46 states.  Oh.  SD had one or two
districts that were SMDs.  MD, NH, VT and WV also had ffa-mmds, but not for
all their seats.  In figures 2 and 3, SD doesn’t appear, so they apparently
caught this mistake later.  So I’m cool with their reported number of
states for state houses (but they should have said they have 46 states for
state houses).

One minor problem with excluding FFA-MMDs that is not dealt with is that a
given locale may change whether it has SMDs or some type of FFA-MMD from
2000 to 2002, or 2010 to 2012, then they are conflating changes over time
with changes in locale.  I’m not checking, but that would conflate change
over time with change in locale.

Their demographic controls are at the state, not the state legislative
district, level.  This isn’t too problematic.  Most of their demographic
controls aren’t necessary, and so are just noise in their model.

Table 5  ; They find that “post-citizens” is associated with both a decline
in the number of Dem candidates and the number of Repub candidates.  Why
would the latter happen?  It’s not a stat sign effect for the latter, but
it disagrees with their theory.

They cluster by state when computing SEs, but is that enough?  They don’t
cluster by state-year.  There should be level-2 errors for state-year, or
they should cluster for state-year.

Please consider hiring me for your state legislative election analysis
needs.

On Sat, Apr 9, 2016 at 10:46 AM, Sean Parnell <
sean at impactpolicymanagement.com> wrote:

> I believe the word you are looking for is actually “evidence,” not “proof.”
>
>
>
> Best,
>
>
>
>
>
> Sean Parnell
>
> President, Impact Policy Management LLC
>
> Alexandria, Virginia
>
> 571-289-1374
>
> sean at impactpolicymanagement.com
>
>
>
>
>
>
>
> *From:* ReThink Media [mailto:tyler at rethinkmedia.org]
> *Sent:* Friday, April 8, 2016 9:41 PM
> *To:* Sean Parnell <sean at impactpolicymanagement.com>
> *Cc:* Rick Hasen <rhasen at law.uci.edu>; law-election at uci.edu
> *Subject:* Re: [EL] “Yes, ‘Citizens United’ gives Republicans an
> electoral edge. Here’s proof.”
>
>
>
> Or just proof that it reduced competition.
>
>
>
> "The decision also made it more difficult to unseat Republican
> officeholders, cementing the already strong financial advantage of
> political incumbents, and reduced the number of Democratic candidates who
> ran for office."
>
> --
>
> Tyler Creighton
>
> *tyler at rethinkmedia.org <tyler at rethinkmedia.org>*
>
>
>
> *Sent from my phone*
>
>
> On Apr 8, 2016, at 11:26 AM, Sean Parnell <sean at impactpolicymanagement.com>
> wrote:
>
> Alternate headline: “Yes, pre-Citizens United independent expenditure bans
> gave Democrats an electoral edge. Here’s proof.”
>
>
>
>
>
> Sean Parnell
>
> President, Impact Policy Management LLC
>
> Alexandria, Virginia
>
> 571-289-1374
>
> sean at impactpolicymanagement.com
>
>
>
>
>
>
>
> Posted in campaign finance <http://electionlawblog.org/?cat=10>, campaigns
> <http://electionlawblog.org/?cat=59>
> “Yes, ‘Citizens United’ gives Republicans an electoral edge. Here’s proof.”
> <http://electionlawblog.org/?p=81684>
>
> Posted on April 8, 2016 7:58 am <http://electionlawblog.org/?p=81684> by *Rick
> Hasen* <http://electionlawblog.org/?author=3>
>
> See this WaPo oped
> <https://www.washingtonpost.com/opinions/yes-citizens-united-gives-republicans-an-electoral-edge-heres-proof/2016/04/07/c9fe3fa4-fb5c-11e5-886f-a037dba38301_story.html>
>  by Tilman Klumpp, Hugo M. Mialon and Michael A. William:
>
> *Unlike the federal government, some states never restricted independent
> political expenditures and were, therefore, unaffected by the Citizens
> United decision. Other states had restricted such expenditures and were
> forced to remove the restrictions after the ruling. In a study that will be
> published in a forthcoming issue of the Journal of Law and Economics, we
> analyzed data from more than 38,000 state legislative races between 2000
> and 2012, in both groups of states. Our objective was to figure out what
> impact, if any,Citizens United had on who gets elected to state legislative
> office. In states that previously banned corporate and union expenditures,
> we found thatCitizens United shifted the odds of electoral success
> detectably and in a clear direction: from Democratic to Republican
> candidates.*
>
> *Many things determine who wins on Election Day, and simple correlations
> don’t automatically indicate causal effects. States that were forced to
> lift their bans on independent expenditures may have elected more
> Republicans in 2010, but this surge could have been caused by a reaction to
> the passage of the Affordable Care Act, the recession or any other issue on
> voters’ minds at the time. And while states with independent expenditure
> bans may have elected more Democrats before 2010 than did states without
> such bans, this does not necessarily have anything to do with how the
> states regulated election finance. There are many reasons some states vote
> differently from others, including historical, cultural and demographic
> differences….*
>
> *With this approach, we found that the chance of Republican candidates
> winning state legislative seats increased by about four percentage points
> on average as a result ofCitizens United, and by 10 or more percentage
> points in several states. The decision also made it more difficult to
> unseat Republican officeholders, cementing the already strong financial
> advantage of political incumbents, and reduced the number of Democratic
> candidates who ran for office. Finally, the data provide evidence
> thatCitizens United discouraged ordinary people from making monetary
> contributions to candidates’ campaigns, an effect feared by critics of the
> decision early on.*
>
> <image001.png>
> <https://www.addtoany.com/share#url=http%3A%2F%2Felectionlawblog.org%2F%3Fp%3D81684&title=%26%238220%3BYes%2C%20%E2%80%98Citizens%20United%E2%80%99%20gives%20Republicans%20an%20electoral%20edge.%20Here%E2%80%99s%20proof.%26%238221%3B&description=>
>
> _______________________________________________
> Law-election mailing list
> Law-election at department-lists.uci.edu
> http://department-lists.uci.edu/mailman/listinfo/law-election
>
>
> _______________________________________________
> Law-election mailing list
> Law-election at department-lists.uci.edu
> http://department-lists.uci.edu/mailman/listinfo/law-election
>

-- 

Dr. Carl Klarner

Academic / Political Consultant

Klarnerpolitics.com

Former Associate Professor of Political Science

Carl.Klarner at gmail.com

Cell: 812-514-9060
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://webshare.law.ucla.edu/Listservs/law-election/attachments/20160409/f409b3d7/attachment.html>