jane green and will jennings university of manchester measuring and analysing mood in party and...
TRANSCRIPT
Jane Green and Will Jennings
University of Manchester
Measuring and Analysing Mood in Party and Government Competence Evaluations in the U.K. and U.S.A.
Nuffield College, March 6, 2012
What should the characteristics of party competence be at the aggregate-level?
What shapes party competence ratings? Do we need macro-competence, if we have partisanship, economic indicators & evaluations, events and PM/presidential approval?
How might the explanation and the consequences of macro-competence vary across institutional settings?
Can - and should – macro-competence be added to our understanding of the macro-polity? 2
Research Questions
Overview
1. Data and Estimation of Mood in Competence
2. Macro-Competence in the UK and USA: validation
3. Modeling Macro-Competence and the Vote (US)
4. Implications and Conclusions
U.K. 1950 – 2008: 2,383 items for party ‘best able to
handle’ and relative handling questions Gallup, Ipsos-Mori, Populus, YouGov, BES.
U.S.A. 1948 – 2010: 2,512 items for party ‘better job’
handling and general performance questions Gallup, AP/Ipsos, ABC/Washington Post, NBC/Wall
Street Journal, YouGov/Polimetrix, NES
Data
7
Part 1 Macro-Competence in the UK and the USA: a validation
U.K. Macro-competence, by party, 1950-2008
54% common variation 63% common variation
20
25
30
35
40
45
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
Conservative Party
20
25
30
35
40
45
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
Labour Party
U.S. Macro-competence, by party, 1944-2010
67% common variation 51% common variation
25
30
35
40
45
50
1950 1960 1970 1980 1990 2000 2010
Republican Party
25
30
35
40
45
50
1950 1960 1970 1980 1990 2000 2010
Democratic Party
U.K.: Macro-competence and Covariates
Macro-Competence Conservative Labour Liberals Government Approval * Governing Party 0.659*** 0.510***
- (118) (118) Leader Ratings 0.796*** 0.790*** 0.411*** (118) (118) (117) Macro-Partisanship 0.562*** 0.804*** 0.606*** (118) (118) (117) Personal Economic Expectations * Governing Party 0.469*** 0.335** 0.053 (118) (118) (117)
* p ≤ .05, ** p ≤ .01, *** p ≤ .001
Note: correlations are tested for interaction of governing party with government approval and personal economic expectations.
U.S. Correlations II: control of the White House
Macro-competence
Governing Party
Opposition Party
Congressional Ballot 0.495*** 0.432***
Macro-Partisanship 0.490*** 0.362*** Consumer Sentiment 0.402*** 0.033***
Presidential Approval 0.779*** -0.618*** Congressional Approval 0.449*** -0.359***
*p≤0.05, **p≤0.01, ***p≤0.001 Start =1980Q1, End = 2009Q4 N = 120
15
Macro-competence and party support, U.K.
25
30
35
40
45
50
55
60
Mac
ro-C
ompe
tenc
e /
Vote
(%)
1979 1983 1987 1991 1995 1999 2003 2007
Macro-Competence Vote Intention
Labour Party
20
25
30
35
40
45
50
Mac
ro-C
ompe
tenc
e /
Vote
(%)
1979 1983 1987 1991 1995 1999 2003 2007
Macro-Competence Vote Intention
Conservative Party
16
Macro-competence and party support, U.S.A.
25
30
35
40
45
50
55
Mac
ro-C
ompe
tenc
e /
Gen
eric
Bal
lot (
%)
1980 1984 1988 1992 1996 2000 2004 2008
Macro-Competence Generic Ballot
Republican Party
25
30
35
40
45
50
55
60
65
Mac
ro-C
ompe
tenc
e /
Gen
eric
Bal
lot (
%)
1980 1984 1988 1992 1996 2000 2004 2008
Macro-Competence Generic Ballot
Democratic Party
Competence and Macro-Partisanship UKGranger-causation between macro-competence and macro-partisanship
Conservative
Party Labour Party
Macro-competence Granger-causes macro-partisanship
χ2test statistic 5.920* 11.081***
p-value 0.052 0.075
Macro-partisanship Granger-causes macro-competence
χ2test statistic 5.164† 1.679
p-value 0.076 0.364
AIC 8.292 7.674
Durbin-Watson d-statistic 1.921 1.789
Lag, selected according to AIC criteria 2 1
N 118 118
* p ≤ .05, ** p ≤ .01, *** p ≤ .001 Start = 1979Q3, End = 2008Q4.
Competence and Macro-Partisanship USGranger-causation between macro-competence and macro-partisanship
Governing
Party Opposition
Party
Macro-competence Granger-causes macro-partisanship
χ2test statistic 8.639* 1.787
p-value 0.013 0.181
Macro-partisanship Granger-causes macro-competence
χ2test statistic 5.751† 1.894
p-value 0.056 0.169
AIC 6.344 6.270
Durbin-Watson d-statistic 2.064 2.167
Lag, selected according to AIC criteria 2 1
N 106 107
* p ≤ .05, ** p ≤ .01, *** p ≤ .001 (variables are pre-whitened through fractional differencing)
Start = 1983Q1, End = 2009Q4. Observations for the period 1980Q1 to 1982Q4 are lost due to fractional
differencing of the variables.
19
Part 2 Macro-Competence in the USA: further validation and analysis
20
Government
Opposition
VOTE
Part
y Co
mpe
tenc
e
Part
y Co
mpe
tenc
e
21
Government
Opposition
VOTE
Part
y Co
mpe
tenc
e
Part
y Co
mpe
tenc
e
Congressional approval by share of party seats
Presidential approval
Macro-partisanship
22
Government
Opposition
VOTE
Part
y Co
mpe
tenc
e
Part
y Co
mpe
tenc
e
Congressional approval by share of party seats
Presidential approval
Macro-partisanship / Public policy mood
Events/Shocks
Economy
Policy Performance
Policy Outcomes
Reputations
Research Questions II: macro-competence US
Is macro-competence responsive to existing measures of performance evaluations (i.e. congressional approval, presidential approval) and macro-partisanship?
Does the interaction of congressional approval and share of the House of Representatives have an effect on macro-competence for the opposition party?
Does governing party competence drive opposition party competence?
What does macro-competence add to models of party support in the US?
Macro-competence for governing and opposition parties
Macro-competence (Gov) Granger-causes macro-competence (Opp)
χ2test statistic 7.575*
p-value 0.023
Macro-competence (Opp) Granger-causes macro-competence (Gov)
χ2test statistic 0.437
p-value 0.804
AIC 7.773
Durbin-Watson d-statistic 2.052
Lag, selected according to AIC criteria 1
N 107
* p ≤ .05, ** p ≤ .01, *** p ≤ .001 (variables are pre-whitened through fractional differencing)
Start = 1983Q1, End = 2009Q4. Observations for the period 1980Q1 to 1982Q4 are lost due to fractional
differencing of the variables.
Fractional persistence of macro-competence
How quickly do shocks persist or dissipate?
If shocks to the reputation of parties due to good or bad performance last indefinitely, public opinion on valence is an integrated process, and if shocks dissipate quickly, it is stationary: determines the degree to which parties are forgiven for past mistakes and mismanagement.
Degree of fractional integration estimated with the Robinson (1995) multivariate semiparametric method, which calculates the value d for macro-competence: equal to 0.56 for the governing party series and 0.71 for the opposition, with the t-statistic indicating these values, as with the covariates, are significantly different from zero.
Fractionally difference the series to avoid spurious estimates.
26
Fractional differencing parameter d
Robinson’s d
d Std Err
t-ratio
d=0
Vote (Gov/Opp) 0.617 0.085 7.279***
Macro-Competence
Governing party 0.559 0.084 6.637***
Opposition party 0.706 0.061 11.515***
Macro-Partisanship (Gov/Opp) 0.772 0.086 9.000***
Presidential Approval 0.756 0.086 8.444***
Consumer Sentiment 0.931 0.083 11.228***
Congressional Approval 0.749 0.080 9.386***
***p<0.001 N = 120, Start = 1980Q1, End = 2009Q4
27
Cointegrating regression: macro-competence and presidential approval
Macro-Competencet
Governing Party
Opposition Party
Presidential Approvalt
0.275*** (0.019)
-0.265*** (0.018)
Carter
-6.037*** (1.599)
-5.526*** (1.507)
Reagan
-7.071*** (1.172)
2.646*** (1.105)
Bush I
-7.308*** (1.232)
4.377*** (1.161)
Clinton
-4.716*** (1.169)
4.605*** (1.102)
Bush II
-5.985*** (1.178)
9.651*** (1.111)
Constant
28.702*** (1.545)
46.193*** (1.456)
Adjusted R-squared 0.720 0.834
Root MSE 2.202 2.076 Robinson’s d of residuals 0.331 0.297
Standard Error 0.113 0.085
T-ratio 2.926*** 3.480*** * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 N = 120; Start = 1980Q1, End = 2009Q4
The model
dMacro-Competence = + dECM (residuals: cointegrating regression)+ dCongressional Approval+ dShare of House of Representatives+ dCongressional approval x House share+ dPresidential Approval+ dMacro-Partisanship+ dConsumer Sentiment
29
Fractional cointegration: macro-competence
∆dMacro-Competencet
Governing Party
Opposition Party
dECMt-1
-0.251*** (0.083)
-0.333*** (0.085)
∆dCongressional Approvalt
1.890† (1.051)
0.965 (1.142)
Share of H of Repst
-0.031 (0.031)
-0.015 (0.024)
∆dCongressional Approvalt * Share of H of Repst
-0.041† (0.024)
-0.016 (0.020)
∆dPresidential Approvalt
0.090** (0.029)
-0.109*** (0.027)
∆dMacro-Partisanshipt
0.740*** (0.172)
0.401* (0.169)
∆dConsumer Sentimentt
-0.026 (0.028)
0.026 (0.027)
Constant
1.003 (1.525)
1.105 (1.224)
Rho 0.311 0.080 Durbin–Watson d-statistic 1.921 1.991
Adjusted R-squared 0.429 0.361
Root MSE 1.418 1.340 † p ≤ 0.1, * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001. N = 108; Start = 1983Q1, End = 2009Q4
Cointegrating regression: macro-partisanship and generic ballot
Votet
Governing Party
Opposition Party
Macro-Partisanshipt
0.702*** (0.033)
0.702*** (0.033)
Constant
13.584*** (1.607)
16.186*** (1.683)
Adjusted R-squared 0.795 0.795
Root MSE 1.917 1.917 Robinson’s d of residuals 0.394 0.394
Standard Error 0.075 0.075
T-ratio 5.268*** 5.268*** * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 N = 120; Start = 1980Q1, End = 2009Q4
The model
dCongressional Ballot = + dECM (residuals: cointegrating regression)
+ dCongressional Approval+ dShare of House of Representatives+ dCongressional Approval x House share+ dPresidential Approval+ dMacro-Partisanship+ dConsumer Sentiment
+ Macro-competence residuals (gov) t-1+ Macro-competence residuals (gov) t+ Macro-competence residuals (opp) t
32
∆dVotet
Governing Party Opposition Party
(1) (2) (3) (1) (2) (3) dECMt-1
-0.271** (0.096)
-0.202* (0.097)
-0.297*** (0.092)
-0.346*** (0.103)
-0.349*** (0.102)
-0.371*** (0.099)
∆dCongressionalApprovalt
0.791 (1.080)
0.710 (1.031)
0.568 (1.008)
-0.510 (1.603)
-0.340 (1.552)
-0.669 (1.526)
Share of H of Repst
-0.093** (0.029)
-0.092*** (0.027)
-0.090** (0.027)
0.017 (0.036)
0.017 (0.035)
0.019 (0.034)
∆dCongressionalApprovalt * Share of H of Repst
-0.019 (0.025)
-0.017 (0.024)
-0.014 (0.023)
0.013 (0.028)
0.010 (0.027)
0.016 (0.027)
∆dPresidentialApprovalt
0.049 (0.032)
0.054† (0.032)
0.059* (0.030)
-0.032 (0.035)
-0.031 (0.033)
-0.046 (0.034)
∆dMacro-Partisanshipt
0.725*** (0.191)
0.735*** (0.188)
0.723*** (0.180)
0.775*** (0.221)
0.805*** (0.213)
0.745*** (0.220)
∆dConsumerSentimentt
0.050 (0.031)
0.048 (0.031)
0.039 (0.031)
-0.040 (0.034)
-0.037 (0.032)
-0.034 (0.033)
∆dMacro-Competence(Gov)(Residuals)t -
0.188† (0.106)
- - - -
∆dMacro-Competence(Gov)(Residuals)t-1 - -
0.205* (0.105)
- - -0.277* (0.119)
∆dMacro-Competence(Opp)(Residuals)t - -
-0.392*** (0.109)
- 0.361** (0.120)
0.340** (0.118)
Constant
6.048*** (1.458)
5.998*** (1.356)
5.893*** (1.354)
1.542 (1.867)
1.532 (1.829)
1.429 (1.782)
Rho 0.148 0.068 0.128 0.300 0.317 0.305 Adjusted R-squared 0.384 0.414 0.468 0.229 0.289 0.310
† p ≤ 0.1, * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 N = 108; Start = 1983Q1, End = 2009Q4
Public opinion on party competence is consistent with a mood about valence in the U.S., (and in the U.K.).
Macro-competence is explained by existing performance variables; these variables (and the ECM) explain between 36% and 43% of variance in macro-competence.
Some evidence that macro-competence is demonstrated in the legislature, for the party in power (as a function of House share).
In a very conservative estimate of the short-run effects of valence, this variable, macro-competence, explains 8-9% of variance in party support (and 3-9% in the U.K.).
Macro-competence is important in the U.S., where it serves as an especially important series for the valence of the opposition; suggesting that the out-party can do relatively little to shape its competence, but its competence is highly important to the vote. 33
Summary
34
Implications
Competence matters over time and in context.
Potential applications of the measure for models of vote choice and electoral forecasting.
Construct of macro-competence is relevant to parties and elites who wish to establish a reputation for competence.
Future questions: Does competence contribute to polarization or moderation
of party positions? Does competence enable parties to widen the issue agenda
of their campaigns? How do governing parties seek to bolster their reputation
for competence when in power? Does the context of policy responsibility mediate the effects
of policy competence?
Jane Green and Will Jennings
University of Manchester
Measuring and Analysing Mood in Party and Government Competence Evaluations in the U.K. and U.S.A.
Nuffield College, March 6, 2012
Economic issues and valence Macro-competence is not a product of economic questions (best
party on the economy); these items compose a small proportion of each measure; including or excluding them makes no substantive difference; economic evaluations load onto the same dimension similarly to other policy issues.
Position issues and valence Ratings on positional issues are still competence-based; they are
not positively associated with public policy mood; they co-vary with other issues; their inclusion or exclusion makes no substantive difference to our results.
Public policy mood and valence We are exploring the dynamics of thermostatic mood for another
paper, but in all our models public policy mood is not simply a function of valence, or vice versa.
Validity considerations
Valence and Macro-Partisanship (USA)Granger Causation Tests between Valence and Macro-partisanship
Republicans Democrats
Valence Granger-causes macro-partisanship
χ2test statistic 7.362* 3.168†
p-value 0.025 0.075
Macro-partisanship Granger-causes Valence
χ2test statistic 3.367 0.825
p-value 0.186 0.364
AIC 6.348 6.321
Durbin-Watson d-statistic 1.97 2.19
Lag, selected according to AIC criteria 2 1
N 106 107
* p ≤ .05, ** p ≤ .01, *** p ≤ .001 (variables are pre-whitened through fractional differencing)
Start = 1983Q1, End = 2009Q4. Observations for the period 1980Q1 to 1982Q4 are lost due to fractional
differencing of the variables.
Autoregressive Distributed Lag Model of Macro-Competence and the Costs of Governing, U.K.
Macro-Competencet Labour Conservative Liberal Macro-Competencet-1 0.867***
(0.062) 0.873*** (0.068)
0.759*** (0.085)
Time in Governmentt
-0.181† (0.105)
-0.140* (0.065)
-
Time in Oppositiont
0.089 (0.058)
-0.001 (0.129)
-
Constant 4.096*
(2.040) 4.436† (2.352)
2.098** (0.767)
* p ≤ .05, ** p ≤ .01, *** p ≤ .001 † ≤ .1
Start: 1950End: 2008
N=59
41
ADL Model of Events on Macro-Competence, U.K.
Start 1979 Q3End 2008 Q4
N=118
Macro-Competencet
Labour Conservative Liberal
Macro-Competencet-1 0.702***
(0.063) 0.885*** (0.095)
0.778*** (0.071)
Macro-Competencet-2 - -0.250**
(0.091) -
Time in Governmentt
-0.030 (0.030)
-0.005 (0.008)
-
Event 1981 Q1: Creation of the SDPt
-2.111* (1.023)
-1.274 (1.036)
1.563* (0.665)
Event 1982 Q3: Falklands Wart
1.946* (0.980)
2.866** (1.007)
-1.267* (0.633)
Event 1984 Q1: Miner’s Striket
-0.054 (0.783)
-1.770* (0.773)
0.505 (0.477)
Event 1988 Q3: Merger of SDP and Liberal Partyt
0.301 (0.793)
0.0168 (0.802)
-0.918 (0.592)
Event 1990 Q1: Poll Taxt
0.341 (1.088)
-0.325 (1.099)
0.071 (0.694)
Event 1991 Q1: Gulf Wart
-1.279 (1.061)
0.873 (1.069)
0.307 (0.687)
Event 1992 Q4: ERM Crisist
2.770*** (0.849)
-2.610** (0.816)
-0.010 (0.433)
Event 2003 Q3: Iraq Wart
-1.828† (0.974)
0.098 (0.529)
0.060 (0.311)
Event 2007 Q4: Northern Rockt
1.196 (0.920)
1.926* (0.905)
0.136 (0.607)
Constant 10.265***
(2.174) 12.174***
(2.230) 2.008** (0.737)
* p ≤ .05, ** p ≤ .01, *** p ≤ .001 † ≤ .1
Start 1979 Q3End 2008 Q4
N=118
∆Macro-Competencet Labour Conservative Liberal Macro-Competencet-1 -0.179†
(0.098) -0.167† (0.096)
-0.240*** (0.055)
Macro-Competencet-2 -0.198*
(0.091) -0.267** (0.087)
-
∆Leader Ratingst
0.060* (0.027)
0.118** (0.040)
0.067*** (0.018)
Leadert-1 0.042*
(0.021) 0.131** (0.042)
0.014† (0.008)
∆Macro-Partisanshipt
0.220* (0.110)
-0.078 (0.084)
0.144 (0.068)
Macro-Partisanshipt-1 0.155†
(0.085) 0.014
(0.045) 0.154*** (0.046)
∆PEXPt
-0.046 (0.034)
-0.089* (0.039)
-0.014 (0.015)
PEXPt-1 -0.008
(0.023) 0.013
(0.019) -0.011 (0.008)
∆GOVt
-6.571 (5.638)
-13.952* (5.633)
-
GOVt-1 3.986
(3.125) -0.180 (3.380)
-
∆(PEXPt*GOVt)
0.096† (0.055)
0.137* (0.055) -
PEXPt-1*GOVt-1 -0.034
(0.033) 0.013
(0.040) -
Time in Governmentt
-0.057† (0.035)
-0.018 (0.011)
-
Constant 6.787†
(4.144) 7.079** (2.368)
1.066 (0.807)
* p ≤ .05, ** p ≤ .01, *** p ≤ .001 † ≤ .1
An Error Correction Model of Macro-Competence, U.K.
Start 1979 Q3End 2008 Q4
N=118
An Error Correction Model of Vote Intention, U.K. ∆Votet
Labour Conservative ECM: Votet-1
-0.372*** (0.06)
-0.376*** (0.070)
-0.451*** (0.078)
-0.479*** (0.081)
∆Macro-Competencet -
0.372* (0.152)
- 0.346*** (0.104)
Macro-Competencet-1 -
0.135 (0.131)
- 0.206* (0.090)
∆Leader Ratingst
0.344*** (0.041)
0.324*** (0.043)
0.341*** (0.047)
0.296*** (0.047)
Leadert-1
0.230*** (0.046)
0.216*** (0.045)
0.169** (0.057)
0.115* (0.056)
∆Macro-Partisanshipt
-0.050 (0.175)
-0.145 (0.178)
0.029 (0.097)
0.040 (0.094)
Macro-Partisanshipt-1
-0.126 (0.129)
-0.179 (0.135)
0.106* (0.052)
0.106 (0.050)
∆PEXPt
-0.141* (0.055)
-0.124* (0.055)
-0.019 (0.045)
0.022 (0.044)
PEXPt-1
-0.182*** (0.043)
-0.179*** (0.042)
-7.924† (6.519)
-0.044* (0.021)
∆GOVt
-16.843† (9.090)
-14.737 (8.969)
-7.924 (6.519)
-2.226 (6.454)
GOVt-1
-24.014*** (5.475)
-25.603*** (5.572)
-9.255 (3.876)
-9.202* (3.734)
∆(PEXPt*GOVt)
0.211* (0.088)
0.178* (0.088)
0.078 (0.064)
0.024 (0.063)
PEXPt-1*GOVt-1
0.254*** (0.057)
0.267*** (0.057)
0.120** (0.046)
0.116** (0.044)
Time in Governmentt
0.051 (0.053)
0.074 (0.057)
-0.027* (0.013)
-0.020 (0.013)
Constant
28.014*** (6.881)
25.570*** (6.956)
8.528*** (2.536)
5.437* (2.710)
0
5
10
15
20
25
30
35
40
45
50
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
Incumbents
0
5
10
15
20
25
30
35
40
45
50
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
Oppositions
Liberals macro-competence
70% common variation in all issue ratings
UK Valence, by party, 1950 - 2008