Partisan Impacts on the Economy:
Evidence from Prediction Markets and Close Elections
and
Party Influence in Congress and the
Economy
Erik
Snowberg
Justin
Wolfers
Eric
Zitzewitz
Stanford GSB
Wharton
Dartmouth
Q Group
October 2007
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Overview
• Do elections matter for asset prices?
• Presidential elections matter. GOP wins =>
– 2% increase in S&P 500
– 10 bp increase in Treasury yields (since 1980)
– Stronger $
• Congressional elections matter less
– GOP control of one house => 20-60 bp increase in S&P
• Broader methodological points
– Prediction markets help calibrate value-relevant news
– Prediction markets can significantly improve on
traditional event studies
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Do elections affect the economy?
• No clear theoretical prediction
– Models predicting policy convergence: Hotelling
(1929), Downs (1957), Lindbeck and Weibull (1987),
Baron (1994)
– Models predicting policy divergence:
Citizen candidate (Besley and Coate 1997), Party
Factions (Roemer 1999), Strategic Extremism (Glaeser
et. al. 2005)
• If they do, which elections matter?
– How important is the Executive Branch vs. Congress?
– How much “control” does control of Congress really
imply?
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Methodological research question
• Can prediction markets “save” event studies?
They help with:
– Measuring pre-event probability
– Determining event window length
– Controlling for other news released with event (e.g.,
impact of 2004 election on 2008)
– Expanding set of analyzable events (non-surprise
events, events that never happen, events before they
happen)
• But require (even greater) faith in efficient markets
• And provide new opportunities for correlation vs.
causation confusion
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Prediction market event studies
Political
Events
Economic
Variables
Re-election
probability
• Forbes and muni bonds (Slemrod and Greimel, 1999)
• UK elections and the FTSE (Herron, 2000)
• A cottage industry prior to the 2004 election
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Issues
• Such studies may be biased:
Fair (1978), Carville’s Law
Economic
Variables
Political
Events
Re-election
probability
Other News
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Two approaches to identification
• Instrumental variables
– An indicator that captures policy news, but not
news that affects outcomes directly
• A market on who will win a debate
• Instrumental events
– An event window in which news is all policy news
– E.g., election night, Dean’s Scream
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
100
Election Night 2004
1150
Exit Polls
Leaked
75
1140
1135
50
1130
Early Returns
Begin
25
0
12:00
15:00
18:00
21:00
0:00
3:00
Time, EST
November 2-3, 2004
% Snowberg,
Probability Bush
wins Presidency
(Tradesports)
S&P 500 Future,
DeliveryInfluence”
Date 12/2004 (CME)
Wolfers,
and Zitzewitz,
“Partisan Effects”
and “Party
1125
1120
1115
6:00
S&P Index Level
Probability Bush wins Presidency, %
1145
Election Night 2004
• To estimate the effect of Bush vs. Kerry we
estimate the model:
Log(S&P 500t) =
Re-election probabilityt+
t
• Prediction markets are slower to move than
financial markets
– use 10 and 30 minute windows
– Scholes-Williams as a robustness check
• Missing observations create heteroskedasticity
– WLS to reduce impact of this
– White (1980) standard errors for robustness to
remaining heteroskedasticity
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Results: 2004
Financial Variable
S&P 500
Dow Jones Industrial
Average
Nasdaq 100
Dollar (vs. Trade-Weighted
Currency Portfolio)
Dec 04
Light Crude
Oil Futures
Dec 05
Dec 06
2 Year T-Note Future
10 Year T-Bill Future
10 Minute Intervals
30 Minute Intervals
Estimate
N
Estimate
N
Dependent Variable:
Log (Price Financial Variable)
0.016***
0.020***
104
35
(0.004)
(0.005)
0.014***
0.016***
84
29
(0.005)
(0.005)
0.019***
0.024***
104
35
(0.005)
(0.008)
0.004
0.005**
84
34
(0.003)
(0.002)
Dependent Variable:
(Price)
1.068***
1.634***
88
29
(0.350)
(0.592)
0.642*
1.032*
85
28
(0.346)
(0.558)
-0.281
-0.901
63
21
(0.751)
(0.782)
Dependent Variable:
(Yield)
0.095*
0.105**
84
30
(0.057)
(0.044)
0.108**
0.116**
91
31
(0.048)
(0.049)
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
How biased is the other method?
Daily
Differences
Financial Variable
S&P 500
Dow Jones Industrial
Average
Nasdaq 100
Dollar (vs. Trade-Weighted
Currency Portfolio)
Light Crude Oil Futures
(Near Month)
10 Year T-Bill Yield
5-day
Differences
20-day
Differences
Dependent Variable:
Log (Price)
0.087***
0.128**
0.243***
(0.034)
(0.062)
(0.065)
0.093***
0.145**
0.275***
(0.032)
(0.064)
(0.090)
0.143***
0.212**
0.299***
(0.062)
(0.098)
(0.107)
0.040**
0.017**
0.021
(0.019)
(0.022)
(0.044)
Dependent Variable:
(Price)
0.468
-7.269**
-12.570***
(4.210)
(3.586)
(3.213)
Dependent Variable:
(Yield)
0.214
0.967*
0.202
(0.299)
(0.523)
(0.598)
In this case, very.
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Results: 2004
• Under Bush v. Kerry:
– Stock indices rise 1.5% - 2.5%
– Oil prices are higher for up to a year
• Long term oil futures not liquid enough for strong conclusions
– The US dollar strengthens somewhat
– Bond yields are 10 basis points higher
• Similar results from inflation-indexed bonds
• Rough interpretation:
– Bush is perceived as good for returns to capital
– But he is perceived as fiscally less conservative than Kerry
– Caveat: this tells us nothing about welfare
• Is this a Republican effect, or an incumbency effect?
– Look at 2000 election
• Is this a Bush effect or a Republican effect?
– Examine earlier elections: 1880 onward
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Election Night 2000
• The 2000 election would make another great
study, except …
– There was no prediction market tracking the
probability of victory of either candidate
• IEM tracked winner of the popular vote
• Centerbet closed their contract the morning of the election
• Centerbet closed with Bush at a 60% chance of
winning
– Can use this to bound the effect of Bush vs. Gore
(the candidates, not the court case)
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Election Night 2000
2:15 Florida
called for Bush
Index, 18:00 November 6 = 100
101
100
4:00 Bush call
rescinded
22:00 Gore call
rescinded
99
98
20:00 Florida
called for Gore
97
18:00
21:00
0:00
3:00
Time (EST) November 6-7, 2000
S&P 500 Future, Delivery date 12/2000 (CME)
Trade Weighted Currency Future, Delivery Date 12/2000 (CME)
Nasdaq 100 Future, Delivery date 12/2000 (CME)
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
6:00
Election Night 2000
Index, 18:00 November 6 = 100
101
100
99
98
97
18:00
21:00
0:00
3:00
Time (EST) November 6-7, 2000
S&P 500 Future, Delivery date 12/2000 (CME)
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
6:00
Election Night 2000
Index, 18:00 November 6 = 100
101
S&P = 0.75% ≥ 1.9%
P(Bush) ≤ 0.4
100
S&P = -0.8% ≥ 1.3%
P(Bush) ≤ -0.6
99
18:00
21:00
0:00
3:00
Time (EST) November 6-7, 2000
S&P 500 Future, Delivery date 12/2000 (CME)
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
6:00
Same exercise in 2004
1150
S&P = 1.3% = 2%
P(Bush) = 0.65
1145
75
1140
1135
50
1130
1125
25
S&P = -0.7% = 2.3%
P(Bush) = -0.3
0
12:00
15:00
18:00
21:00
0:00
1120
3:00
Time, EST
November 2-3, 2004
% Probability Bush wins Presidency (Tradesports)
S&P 500 Future, Delivery Date 12/2004 (CME)
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
1115
6:00
S&P Index Level
Probability Bush wins Presidency, %
100
Results: 2000 and 2004
2000
2004
Financial Variable
21:00
2:15
S&P 500
2.0%
1.3%
1.9%
Nasdaq 100
2.4%
3.7%
3.0%
Dollar (vs. Trade-Weighted
Currency Portfolio)
0.5%
0.6%
0.7%
• Estimates in-line with each other and 2004
-> a Bush effect, not an incumbency effect
• Can we say more about partisanship’s
effect on these economic variables?
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
How Common are Such Events?
Somewhat.
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Electoral Surprise
• Electoral Surprise is defined by:
I(Rep. won) - P(Rep. win before election)
• If, before an election the Republican has a 60%
chance of winning then electoral surprise is:
– + 40% if the Republican wins
– - 60% if the Democrat wins
• Every election has some degree of surprise,
ranging from:
– Slight (3%) in Davis vs. Coolidge (1924)
– Extreme (-89%) in Truman vs. Dewey (1948)
Markets move in response to this surprise
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Pre-election close to Post-election close
Equities and Electoral Surprise
Equity Market Responses to Electoral Shocks: 1880-2004
.04
McKinley-1896
Clinton-1996
McKinley-1900
Taft-1908
Wilson-1912
Reagan-1980
Roosevelt-1936
Coolidge-1924
Hoover-1928
Roosevelt-1904
Harding-1920
.02
GWBush-2004
Kennedy-1960
Harrison-1888
Eisenhower-1952
Reagan-1984
Cleveland-1892
Nixon-1968
Johnson-1964
Roosevelt-1944
Bush-1988
Wilson-1916
Nixon-1972
Garfield-1880
0
Clinton-1992 Eisenhower-1956
Carter-1976
Cleveland-1884
GWBush-2000
-.02
Roosevelt-1940
-.04
Roosevelt-1932
Truman-1948
-1
-.75
-.5
-.25
0
.25
.5
.75
Shock to Probability of a Republican President
I(Republican President) - Pre-election prediction market price
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
1
Equities: 1880 - 2004
Prob(GOP President)
Prob(Incumbent)
• Santa-Clara and Valkanov (2003) conduct a
similar analysis and find no election day effects
(1928-1996)
– In their analysis shock is always ±100%
• Longer-run analysis confirms incumbency is not
a factor
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Pre-election close to Post-election clos
Bonds and surprises
Bond Market Responses to Electoral Shocks
Pre-Reagan: 1892-1976
15
Since Reagan: 1980-2004
Reag an -19 80
15
Reag an -19 84
10
10
Carter-1 97 6
5
GWBush-200 4
5
Roo sev elt-19 32
Nixo n-1 97 2
Coo lidg e-192 4
0
Clin ton -19 92
Wilson -19 16
Roo sev
Wilson
Jo
elt-19
hnHard
so
-19
44
Eisen
McKinley-19
Hoo
n-1
Taft-190
12
ing96
ver-19
-1
ho4920
wer-1
Eisen
8 2895
00ho
6 wer-1 95 2
Roo sev elt-19 04
Truman -19 48
Nixo n-1 96 8
Clev eland -18 92
Kenn ed y-1 96 0
Bush -1
GWBush-200
98 8
0
0
Roo
Roo
sev
sev
elt-19
elt-19
4036
-5
-5
-1
-.75 -.5
-.25
0
.25
.5
.75
1
Clin ton -19 96
-1
-.75 -.5
-.25
0
.25
.5
.75
1
Shock to Probability of a Republican President
I(Republican President) - Pre-election prediction market price
Change in bond yields = 0.03 + 1.48*Post_1980 + Change in prob(Republican)*(0.05 + 12.79*Post_1980)
(0.91) (2.31)
(2.60) (8.51) R-sq=0.23
Note: Chart shows estimate of bond market reaction for 2004 estimated from Table 1
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Other issues
• 2008 Presidential party odds did not change
– Suggests no incumbent party advantage for Presidency
• Adding Congressional control contracts in 2004 does not
alter results
– Not much variation in Congressional probabilities on election day
– Next paper suggests effect of Congress not that large anyway
• Adding controls for expected electoral vote total in 2004
does not alter results
– Suggests winner mattered much more than margin of victory
• Do markets rally in response to the resolution of
uncertainty?
– No evidence of this
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Other issues (2)
• Did market declines when exit polls leaked influence late
voting?
– Maybe, but similar estimates from 3-6 pm and 6pm-midnight
windows suggests this is not biasing results
– Comparison of exit poll panels suggests no shift toward Bush in
evening vote
• Reconciliation with studies showing higher stock returns
under Democrats (e.g., Santa Clara and Valkanov, 2003)
– These studies have low power (SD of 4-year returns = 37%,
N since 1880 = 31 => SE of GOP-Dem difference = +/- 10%)
– Excluding 1929-36 flips sign
– Results suggest that traders do not expect past outperformance
under Dems to continue
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Party Influence in Congress
and the Economy
Erik Snowberg
Stanford GSB
Justin Wolfers
Wharton – U.Penn
CEPR, IZA & NBER
Eric Zitzewitz
Stanford GSB
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
What about Congress?
• Competing theories of legislator behavior
– Cartel theories (Cox and McCubbins, 2002)
• Policy is made by the median member of the majority
– Suggests similar effects as in Presidential elections
(Median Democrat v. Median Republican)
– Weak discipline (Krehbiel, 1993)
• Party labels are only a rough description of preferences
– Suggests policy made by median representative
– Median rarely shifts much
(Joe Lieberman v. Lincoln Chafee)
• “Divided” versus “Unified” Government
– Is divided better than unified control by either party?
– Was the Democrats winning 2 houses in 2006 less than
twice as important as their winning one?
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
0
20
40
60
80
100
GOP victory probabilities every 10 minutes, 11/7/2006
18
20
22
Eastern Time
HOUSE
VA
MO
24
SENATE
MT
TN
Bid-ask midpoints, data treated as missing if bid-ask spread is wider than 10 points.
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
26
Close-up: 2006 Senate experiment 9 p.m.-2 a.m.
100
Magnifying y-scale for for S&P 500
100.00
2006 Senate Race
99.95
S&P 500: Index 4pm=100
80
60
40
99.90
99.85
20
99.80
0
99.75
9 p.m.
10 p.m.
11 p.m.
Election Day
Midnight
1 a.m.
2 a.m.
Post-election Day
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Congress v. The White House
Dependent Variable:
S&P 500
Pr(Senate)
Pr(House)
n
Pr(President)
n
0.174*
(.089)
0.537
(.729)
30
2.05***
(.503)
35
Table 1
Log(Price)/100
Price ($)
Nasdaq
Japanese
Canadian
Euro
Oil
100
Yen
Dollar
Congressional Elections, 2006
0.156
0.025
0.087
0.022
0.034
(.101)
(.045)
(.061)
(.045)
(.079)
-0.017
0.015
-0.450
0.553
-0.209
(1.11)
(.707)
(.634)
(.687)
(.893)
29
28
28
28
30
Presidential Elections, 2004
***
*
2.40
-0.531
-0.694* -0.488**
1.706**
(.814)
(.296)
(.358)
(.221)
(.659)
35
34
34
34
29
Yield (bp)
2 Year
10 Year
T-Note
T-Note
0.918
(1.15)
-1.31
(6.54)
26
0.701
(.693)
-2.11
(6.09)
27
10.8***
(3.58)
30
12.0***
(4.65)
31
Notes: ***, **, * denote statistically significant at 1%, 5% and 10%, respectively. (White standard errors in parentheses.)
Sample period is 4 p.m. Eastern Time on 11/7/2006 to nine a.m. on 11/8/2006, for 2002 it is 4pm on 11/5/2002 to 12pm on 11/6/2002.
Election probabilities are the most recent transaction prices collected every thirty minutes from Tradesports.com. When there are
missing observation, and the bid and ask prices are within 10 percentage points of each other, we use the bid-ask average.
All futures have delivery dates of December 2006, December 2004 and December 2002, respectively.
• Supplementary evidence: Large “Rumsfeld effect”
– S&P 500 rose 0.6% on news of his resignation
– Tradesports suggests that markets had already priced in a 40%
chance of a 2006 resignation
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Election Night: 2002
100
100.9
S&P 500: 4 p.m. on Election Day = 100
Election Night: 2002
80
60
40
100.6
100.3
100.0
20
4 p.m.
99.7
6 p.m.
8 p.m.
10 p.m.
Midnight
Probability of a Republican Majority in the Senate (Tradesports)
Probability of a Republican Majority in the House (Tradesports)
S&P 500 Future (CME), right axis
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Congressional Elections, Compared
Dependent Variable:
S&P 500
Pr(Senate)
Pr(House)
n
Pr(Senate)
0.593**
(.231)
13
Pr(President)
2.05***
(.503)
35
n
n
0.174*
(.089)
0.537
(.729)
30
Table 1
Log(Price)/100
Price ($)
Nasdaq
Japanese
Canadian
Euro
Oil
100
Yen
Dollar
Panel A: Congressional Elections, 2006
0.156
0.025
0.087
0.022
0.034
(.101)
(.045)
(.061)
(.045)
(.079)
-0.017
0.015
-0.450
0.553
-0.209
(1.11)
(.707)
(.634)
(.687)
(.893)
29
28
28
28
30
Panel B: Congressional Elections, 2002
0.728
-0.196
0.011
0.071
0.148**
(.419)
(0.187)
(.127)
(.068)
(0.060)
12
13
13
12
14
Panel C: Presidential Elections, 2004
***
2.40
-0.531*
-0.694* -0.488**
1.706**
(.814)
(.296)
(.358)
(.221)
(.659)
35
34
34
34
29
Yield (bp)
2 Year
10 Year
T-Note
T-Note
0.918
(1.15)
-1.31
(6.54)
26
0.701
(.693)
-2.11
(6.09)
27
4.51*
(1.81)
7
2.21*
(1.02)
7
10.8***
(3.58)
30
12.0***
(4.65)
31
Notes: ***, **, * denote statistically significant at 1%, 5% and 10%, respectively. (White standard errors in parentheses.)
Sample period is 4 p.m. Eastern Time on 11/7/2006 to nine a.m. on 11/8/2006, for 2002 it is 4pm on 11/5/2002 to 12pm on 11/6/2002.
Election probabilities are the most recent transaction prices collected every thirty minutes from Tradesports.com. When there are
missing observation, and the bid and ask prices are within 10 percentage points of each other, we use the bid-ask average.
All futures have delivery dates of December 2006, December 2004 and December 2002, respectively.
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
100.5
Election Night: 1994 Republican Takeover
1994
Election: Republican
Takeover
100.4
100.3
CBS calls House for GOP
CNN calls House for GOP
TV anchors hint at pro-GOP exit polls
NPR reports GOP win >=6 Senate seats
CNN calls Senate for GOP
100.2
100.1
100.0
Early exit polls available to party operatives
4 p.m.
6 p.m.
8 p.m.
10 p.m.
Midnight
S&P 500 Future (CME)
Note: Timing of info reflects press reports. (Intraday prediction markets were unavailable to track electoral shocks.)
Election eve: Prob. GOP House majority = 20%; Prob. GOP Senate majority = 70%
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
1994 Republican Takeover
• Election-eve odds:
– Prob.(GOP Senate) = 70%
– Prob.(GOP House) = 20%
• From 4pm-11pm
–
–
–
–
–
S&P 500 rose 0.40% (standard error = 0.29%)
Oil prices rose $0.26 (0.07)
Yen rose 0.17% (0.18)
German Mark rose 0.20% (0.17)
10-year Treasury note was unchanged
• Standard errors are SD of 4pm-11pm changes
from 50 weekdays prior to 1994 election and 50
weekdays after.
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Jeffords Experiment
• May 2001: Jim Jeffords handed the senate
back to the Democrats
– Jayachandran (JLE 2006) estimates
• Average Democrat-supporting firm rose 0.11%
• Average Republican-supporting firm fell 0.43%
• Net effect = -0.32%
– Unfortunately the Jeffords defection played out
over too long of a period (a week) for us to
apply our methods
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Divided Govt v. Unified Govt
• “The prospect of gridlock has been welcomed, as usual,
on Wall Street”
– John Tierney, NY Times, 11/14/2006
• 2006: Republican House loss created divided govt
– Yet markets fell in response
• 2002: Republican Senate victory created unified govt
– Yet markets rose in response
• 1998: No experiment: Republicans won as expected
• 1994: Republican takeover
– Markets rose, but roughly consistent with pure
partisan effect
• Earlier Presidential elections
– On average no effect of unified goverment
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Shape of the stock marketGOP control relationship
S&P 500 (M)
2006 House: 54bp
2002 Senate: 59bp
GOP WH,
2 GOP Hses
2006 Senate: 17bp
GOP WH,
1 GOP Hse
2004 Pres: 200bp
2000 Pres: >160bp
GOP WH,
0 GOP Hses
1994 Congress:
73bp
Dem WH,
2 GOP Hses
No evidence of Tierney’s story (M’’(c)<0)
Dem WH,
0 GOP Hses
Some evidence for M’’’(c)>0
Degree of GOP control (c)
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Conclusions about substance
• Evidence of systematic partisan effects
– Effects on equity markets consistent with usual characterization
of right-wing parties
– But bond market effects differ: Republicans cause higher
interest rates.
• Congressional elections yield:
– Similar qualitative pattern of partisan effects
– But quantitatively much smaller
– Suggesting weak party discipline (or Congress irrelevant)
• Partisan effects overwhelm any preference for divided
versus unified government
• Election-night experiments yielded very different
estimates than previous observational studies
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”
Conclusions about method
• Prediction markets can help solve many of the
problems with traditional event studies
– Measuring pre-event probabilities
– Identifying event window
– Increasing precision (our SE for Senate 2006 is
0.09% vs. 0.9% for 1928-96 using traditional method)
– Controlling for other information released
• But one needs to be very careful about
identification
– Pre-election correlations were very misleading
Snowberg, Wolfers, and Zitzewitz, “Partisan Effects” and “Party Influence”