Vous êtes sur la page 1sur 24

Munich Personal RePEc Archive

Prediction for the 2012 United States


Presidential Election using Multiple
Regression Model
Pankaj Sinha and Aastha Sharma and Harsh Vardhan Singh
Faculty of Management Studies, University of Delhi
5. August 2012
Online at http://mpra.ub.uni-muenchen.de/41486/
MPRA Paper No. 41486, posted 22. September 2012 06:25 UTC
1

Prediction for the 2012 United States Presidential Election
using Multiple Regression Model
Pankaj Sinha, Aastha Sharma and Harsh Vardhan Singh
Faculty of Management Studies, University of Delhi
Abstract
This paper investigates the factors responsible for predicting 2012 US !residential election Though
contemporary discussions on !residential election mention that unemployment rate "ill be a deciding
factor in this election, it is found that unemployment rate is not significant for predicting the
forthcoming !residential election #$cept %D! gro"th rate, various other economic factors li&e interest
rate, inflation, public debt, change in oil and gold prices, budget deficit'surplus and e$change rate are
also not significant for predicting the US !residential election outcome (e"is)*ec& and +ice ,1-.2/
proposed %allup rating, obtained in 0une of the election year, as a significant indicator for forecasting
the !residential election 1o"ever, the present study finds that even though there e$ists a relationship
bet"een 0une %allup rating and incumbent vote share in the !residential election, the %allup rating
cannot be used as the sole indicator of the !residential elections 2arious other non)economic factors
li&e scandals lin&ed to the incumbent !resident and the performance of the t"o parties in the midterm
elections are found to be significant 3e study the influence of the above economic and non)economic
variables on voting behavior in US !residential elections and develop a suitable regression model for
predicting the 2012 US !residential election The emergence of ne" non)economic factors reflects the
changing dynamics of US !residential election outcomes The proposed model forecasts that the
Democrat candidate Mr *arac& 4bama is li&ely to get a vote percentage bet"een 51.1. 6 ) 5728- 6,
"ith -56 confidence interval
1. Introduction
The outcome of the United States !residential elections has a significant influence on "orld economies,
developed or developing The road to the 3hite 1ouse is a roc&y and long one 4ver the years, several
researchers have tried to e$plain the results of the electoral contest "hich leads to one of the most
po"erful 9obs in the "orld, that of the !resident of the United States of :merica
2

Several models have been proposed in the past for forecasting !residential elections result (iterature
suggests that economic variables can be ta&en as the independent variables "hile some suggest use of
non)economic variables to get the forecasts #ach method has its respective merits and de)merits ;n
this paper, "e e$plore the impact of various economic and non)economic variables on the !residential
forecasting model
For prediction of !residential elections, researchers such as Fair ,1-<., 2012/ analy=e economic factors
such as ,annual rate/ gro"th rate of real per capita %D! in the first three >uarters of the election year
:bramo"it= ,1-../ uses the gro"th rate of the economy in the first si$ months of the election year
(ichtman ,2005, 200./ also refers to the gro"th rate to develop t"o of the ?&eys@ for the 3hite 1ouse
Sinha and *ansal ,200./ derive predictive density function under 1ierarchical priors and use these
results to forecast 200. US !residential election using +ay FairAs model
:part from gro"th, inflation is the second variable that is "idely accepted as an indicator of the
economic health of a country Fair ,1-<., 2012/ used the absolute value of the gro"th rate of the %D!
deflator as an indicator to gauge the election results The same definition is used by Bu=an, 1eggen and
*undric& ,2000/ to analyse the outcome of presidential elections outcomes based on simulation run
over fiscal models
The third macroeconomic variable in consideration is unemployment rate of the United States
;ntuitively spea&ing, several researchers might consider the unemployment rate to be a measure of the
discontent of the people The change in unemployment rate has also been used to forecast election
results by researchers including 0CrDme and 0CrDme )Spe=iari ,2011/ 1o"ever, the ine$act nature of this
relationship has been highlighted by Silver ,2011/, finding that there has been no relationship bet"een
the unemployment rate and the margin of victory ,defeat/
So, the ma9or macroeconomic variables "hich have been studied to influence !residential election
outcomes are the gro"th rate, the unemployment rate and the inflation rate of an economy
Several studies have chosen to loo& at economic factors in a different manner altogether #ri&son and
3le=ien ,1--E/ chose to refer to economic indicators holistically, loo&ing at the inde$ of leading
economic indicators The *read and !eace model by 1ibbs ,2000, 2012/ considers gro"th in real
disposable per capita income as an economic indicator to measure the li&elihood of the incumbent party
in an election to retain the 3hite 1ouse
8

#ven though the &ey indicator,s/ to any !residential #lection are "idely accepted to be economic in
nature, non)economic factors do play a ma9or role in the determination of an election outcome The
most prominent non)economic factor in this regard is "hether the country is currently involved in any
military interventions 3ar'peace have been referenced as a factor in studies done by Fair ,1-<., 2012/,
1ibbs ,2000,2012/, (ichtman and Feilis)*oro& ,1--E/
:nother &ey non economic factor "hich emerges from past discussions in the field is that of
incumbency Got 9ust "ith regard to the incumbent !resident running for office again, but also the
number of terms the incumbent party has spent in office plays a role in the re)election prospects Fair
,1-<., 200./, *artells and Haller ,2001/ and (ichtman and Feilis)*oro& ,1--E/ refer to incumbency as a
factor for reelection :lan ; :bramo"it= ,1-../ adds to the research in the field by building a model that
included a ?time for change@ factor) dependent on the number of terms the incumbent party has been
in po"er
(e"is)*ec& and +ice ,1-.2/ include presidential popularity as a factor to formulate a model for
predicting the result of the presidential elections The inclusion of these factors comes from e$tending
(ee SeigelmanIs ,1-<-/ "or& "hich proves that there e$ists a relationship bet"een the popularity rating
of the incumbent president and the preceding election SeigelmanAs model provides a relationship
bet"een the popular vote share of the incumbent and the %allup rating as obtained on the last pre)
election popularity poll (e"is)*ec& et al ,1-.2/ ho"ever, proposes the use of %allup +atings as obtained
in 0une of the election year This decision to ta&e the 0une rating rather than a rating closer to the
process has multiple reasons, the primary one being that the closer the process is, the larger the
electoral mood s"ings are li&ely to be :s discussed by (e"is)*ec& et al ,1-.2/, the 0une rating is ideal
since it measures 9ob approval in a period of relative political calm, pre)conventions and post)primaries
Though in the present study "e find that there e$ists a relationship bet"een %allup ratings ,0une/ and
!residential #lection, it is "orth noting that the %allup survey, though e$tensive, is non)e$haustive in
nature, and hence canAt be used as the sole indicator of the !residential #lections
:part from the presidential elections, the other Federal elections held in the US: consist of those for the
Bongress) ie the Senate and the 1ouse of +epresentatives 3hile the Members of the 1ouse of
+epresentatives ,often called the 1ouse/ have a term of t"o years, Members of the Senate have a si$
year long term, staggered such that 886 of the Senate undergoes elections every t"o years ;n almost
all of the midterm elections held since 1-7., the incumbent party has made gains in the midterm 1ouse
elections only in the 1--. and 2002 elections The midterm 1ouse election is typically seen as a
7

referendum on the ruling party ,Tufte, 1-<5/ :s discussed by Tufte ,1-<5/, it is almost inevitable that
the incumbent party "ill lose seats in the 1ouse, the important thing to note becomes "hether or not it
"ill lose the ma9ority sta&e after the midterm elections
;n vie" of the above studies, it seems that using a combination of economic and non economic factors, a
regression model can be used for predicting US !residential elections ;n this paper, "e study the
influence of various economic and non)economic variables on voting behavior in US !residential
elections and develop a suitable regression model for predicting the 2012 US !residential election
The section 2 of this paper lists out all the factors that "e consider during the development of the
prediction model ;t also discusses the sources of the data used in this research Section 8 analy=es
various regression models using economic and non)economic variables 2ariables for the proposed
model are selected after a careful analysis ;n section 7, "e test the proposed model by forecasting 200.
US !residential election Section 5 presents the election forecasts of the 2012 election using the
proposed model
2. Significance of Factors onsidered
;n this section, on the basis of the above literature revie", "e analy=e various factors ,economic and
non)economic/ to find out the significant variables that could be used in forecasting !residential
elections The follo"ing factors, listed under the t"o categories, have been consideredJ
Economic Factors: ;n this, "e discuss the different economic factors that might affect the US
!residential election outcome 3hile factors li&e unemployment rate, gro"th rate, inflation, interest
rate and healthcare budget affect the perception of the citi=ens about the incumbent !residentAs "or&
and effectively that of the ruling party, other factors li&e budgetary deficit'surplus and public debt signal
the robustness of the national economy : budgetary deficit is a signal of economic overspending and
could turn out to be favorable for the incumbent party 4n the other hand, a budgetary surplus is a
signal of conservative management of the economy and might be unapproved by the citi=ens %lobal
economic indicators li&e oil prices, gold rate and e$change rate reflect the impact of the state of the
foreign economies on the United States economy and might have an impact on the election results The
follo"ing is the list of the various economic factors consideredJ
1 Unemployment +ateJ :nnual average unemployment rate ,percent/ of civilian labor force
ie 1E years and over, as defined, by *ureau of (abor Statistics ,2012a/ To further e$plore
5

the impact of the national unemployment rate on the !residential results, the >uadrennial
percentage change in unemployment rate is also considered
2 *udget Deficit'SurplusJ The e$cess or deficit of budgetary receipts over the budgetary
outlays as a percent of %D!, as defined by The 3hite 1ouse ,2012/
8 %old !ricesJ The >uadrennial ,6/ change in annual average price per ounce of %old
7 #$change +ateJ The >uadrennial change in the e$change rate of K "ith Franc, #uro, Mar&
and !ound is considered For illustration, K'L e$change rate is used
5 4il !ricesJ The >uadrennial ,6/ change in average crude oil price in dollars per barrel
E ;nterest +ateJ The annual Federal funds effective rate as mentioned by the Federal +eserve
,2012/
< !ublic DebtJ %overnment debt as a percentage of the nominal %D! as defined by
;nternational Monetary Fund ,2010/
. %ro"th +ateJ The gro"th rate of the real per capita %D! in the first three >uarters of the
election year ,annual rate/ as defined by Fair ,200E/
- ;nflationJ The absolute value of the gro"th rate of the %D! deflator in the first 15 >uarters
of the administration of the incumbent president ,annual rate/ as defined by Fair ,200E/
10 1ealthcare *udgetJ The social benefit spending by the Federal and the State government as
a percentage of the national %D!, as provided by *ureau of #conomic :nalysis ,2012/
The data for the economic variables is summari=ed in Tables 5a and 5b given in the :ppendi$
Non-Economic Factors: :s "e shall observe later in this study, it is misleading to assume that US
!residential election is a product of only economic factors : lot of social and non)economic factors have
significant influence on the election outcome The !residential "or& approval rating is a reflection of the
perception of the citi=ens about the "or& done in the incumbent term 4ther non)economic factors
listed belo" might be a source of information about the citi=ensA opinion of the incumbent partyAs
credentials in the electionJ
1 !residential 3or& :pproval +atingJ !ercentage of the :merican population that approves or
disapproves of the "or& done by the incumbent !resident #ven though it is easy to identify
from the contemporary literature that the %allup 9ob approval rating is the most reliable and
"idely accepted measure, the other considered !residential 9ob approval rating are +eal
Blear !olitics ,since 2000/, +asmussen +eports ,since 2008/, BGG'4+B ;nternational Survey
,since 1-.0/, :ssociated !ress)%fF ,since 200./ :s "e observe, most of these ratings are
E

recent and not suitable for our analysis The data used in the analysis is the average %allup
rating of the incumbent !resident in the month of 0une of the election year :lso, the
average rating of the incumbent !resident during the tenure is studied ;n situations "hen
the elected !resident resigns or passes a"ay, the approval rating of the incumbent presiding
over the current election is considered
2 ;ncumbent !resident ScandalsJ ;t is the severity of a scandal that occurred in the term of an
incumbent !resident Scandals could be political, personal or of any other &ind that
negatively affect the incumbent !residentAs popularity and conse>uentially that of the
incumbent party The ratings ,0)2/ for each of the !residential terms have been given as
follo"sJ
Go ma9or scandal in the !residential termJ 0
:t least one ma9or scandal of some severity in the !residential termJ 1
Scandal of high severity, leading to the possible impeachment'resignation of the
incumbentJ 2
Table 1 lists the above ratings for the various incumbent !residents These values are
chosen after detailed ne"s and literature analysis :s "e observe, only in the case of
!resident Gi$on ,1-<E/ and Blinton ,2000/, the ma$imum rating of 2 has been given
8 Military ;nterventionsJ ;mpact of the military interventions during the incumbent rule on
the perception of the incumbent party The ratings ,)1 to M1/ for each of the !residential
terms have been given as follo"sJ
The intervention during the !residential term improved the incumbent
popularityJ M1
The intervention during the !residential term had no impact on the incumbent
popularityJ 0
The intervention during the !residential term sabotaged the incumbent
popularityJ )1
For e$ample, "hile Mr %eorge 3 *ush gets a )1 rating for his second !residential tenure
due to the unpopular ;ra> and :fghanistan "ars, Mr *arac& 4bama gets a M1 for bringing
the "ar to an end
<

7 Midterm !erformanceJ :s discussed earlier, it is observed that the incumbent partyAs
performance usually dips in the midterm elections 1o"ever, the degree of this
performance is a direct indicator and a referendum on the incumbent partyAs popularity
after 2 years in office 3hile the previously mentioned !residential 3or& :pproval +ating is
influenced by the personality and other features of the incumbent !resident, the midterm is
a clearer indicator of the incumbent partyAs acceptability The variable midterm is calculated
as follo"sJ
For election year InAJ

(|E0uSE
SLA1S
] E0uSE
RLS0L1
+|SENAIE
SLA1S
] SENAIE
RLS0L1
)
E0uSE
SLA1S
+SENAIE
SLA1S


"here)
HOUSE
SEATS
J Total number of seats in the 1ouse occupied by the Democrat and +epublican
party representatives during the midterm election prior to the forthcoming election InA
SENATE
SEATS
J Total number of seats in the Senate occupied by the Democrat and +epublican
party senators during the midterm election prior to the forthcoming election InA
The variable HOUSE
RESULT
ta&es value bet"een ,)1 and 1/ as follo"sJ
1 if the incumbent party is in ma9ority in the 1ouse after the midterm election
prior to the forthcoming !residential election InA
)1 if the incumbent party is in minority in the 1ouse after the midterm election
prior to the forthcoming !residential election InA
0 other"ise,
The variable SENATE
RESULT
ta&es value bet"een ,)1 and 1/ as follo"sJ
1 if the incumbent party is in ma9ority in the Senate after the midterm election
prior to the forthcoming !residential election InA
)1if the incumbent party is in minority in the Senate after the midterm election
prior to the forthcoming !residential election InA
0 other"ise
4n an average, the number of seats in 1ouse is close to 7 times of the seats in Senate
1ence, the variable is tilted to"ards the values of 1ouse
Seats
+ather than being a fla", it is
useful in being a better estimator of the incumbent partyAs public perception and acts as a
.

measure of referendum This is due to the fact that in midterm elections, all the seats of the
1ouse go to re)election "here as in Senate, on an averageN only 886 of the seats are being
contested for re)election
For e$ample, for the election of 200., the value for 1ouse
+esult
O)1 since the number of the
incumbent +epublican partyAs seats in 1ouse ,202/ is less than the number of seats of the
Democrat ,288/ Similarly, Senate
+esult
O 0 as the number of +epublican and Democratic seats
are e>ual after the midterm election ,7- each/

The data for the economic variables is summari=ed in Tables 1, 2, 8 and 7 given in the :ppendi$
!ata Sources
Since the %allup rating for the !residents elected before 1-7. is not available, all the values for the
economic and non)economic variables have been considered since 1-7. only The gro"th and inflation
rate are referred from Fair ,200E, 200., 2012/ The unemployment rate is retrieved from the *ureau of
(abor Statistics ,2012b/ The budgetary surplus'deficit data has been ta&en from The 3hite 1ouse
,2012/ 1istorical gold prices are ta&en from the United States Gational Mining :ssociation ,2011/ and
the K'L e$change rate has been obtained from the *an& of #ngland ,2010/ 1istorical oil prices have
been ta&en from ;nflationDatacom ,2012/ The historical Federal funds rate has been obtained from the
Federal +eserve ,2012/ and the healthcare e$penditures data is found at *ureau of #conomic :nalysis
,2012/ The data on public debt has been obtained from ;nternational Monetary Fund ,2010/
Gon)economic factors li&e scandals and military interventions have been arrived at after revie"ing the
contemporary literature on the past !residential tenures These include the articles and essays
mentioned on the history of United States president li&e the dedicated 3hite 1ouse resource and other
reliable lin&s li&e Miller Benter The historical %allup average rating in 0une of the #lection Pear and
:verage %allup term rating "ere obtained from the %allup !residential !oll ,2012/ 3hile the historical
data "as available for most of the independent factors, the Federal effective interest rate "as found
only since 1-5E The results for the historical Bongress elections have been collected from the 4ffice of
the Bler& ,2010/ :lso, the >uadrennial change in oil prices "as available 1-52 on"ards
The dependent factor in our analysis is the vote percentage of the incumbent party in the t"o)party
!residential election as given in Fair ,200E, 200./
-

". Methodolog#
#conomic FactorsJ The follo"ing table lists the regression results for various models containing
economic variables as independent variables and VOTE as dependent variableJ
Table :J :nalysis of ;nfluence of #conomic 2ariables
Model Pear +
2
,6/ p)Statistic
I0IE =
1
+
2
0R0wIE +
3
INFIAII0N
+
4
uNEHPI0HENI_RAIE
1-7.)
200.
88<<5
%+43T1
Q

;GF(:T;4G
UG#M!(4PM#GTR+:T#
007E
0551
0<5E
I0IE =
1
+
2
0R0wIE
+
3
CEAN0E_uNEHPI0HENI_RAIE
1-52)
200.
81-7.
%+43T1
Q

B1:G%#RUG#M!(4PM#GTR+:T#
008E

0-08
I0IE =
1
+
2
0R0wIE +
3
EEAIIECARE +
4
EBI
1-7.)
200.
8<508
%+43T1
Q

1#:(T1B:+#
D#*T
0070
085-
05.1
I0IE =
1
+
2
0R0wIE +
3
INIERESI_RAIE
1-5E)
200.
808E.
%+43T1
Q

;GT#+#STR+:T#
0057
0--2
I0IE =
1
+
2
0R0wIE +
3
00I +
4
0II
1-52)
200.
78-2.
%+43T1
Q

%4(D
4;(
00<0
057-
01<5
I0IE =
1
+
2
0R0wIE +
3
EFICII
+
4
EXCEAN0E_RAIE
1-7.)
200.
85..2
%+43T1
Q

D#F;B;T
#SB1:G%#R+:T#
0025
0750
0.7.
*Denotes signifiant !"#a$%e at & ' $e#e$ of signifiane
The analysis suggests that several economic variables are not able to predict the !residential election
results effectively Unemployment rate and the >uadrennial change in unemployment rate are
insignificant #$ternal factors li&e percentage change in oil price, gold price and Sterling e$change rate
have been also found insignificant ;nternal monetary and budget factors li&e interest rate, public debt
and budget deficit'surplus also do not affect vote percentage of !residential outcomes
%D! gro"th rate is the only important significant variable in the above regression models This is in
contrast to the "idely held belief in the contemporary literature that the forthcoming US !residential
election "ill be decided on economic factors such as unemployment rate, inflation, budgetary deficit
and public debt
Gon)economic Factors
*esides the already defined %allup T :verage)%allup +ating, Scandals, 3ars and MidtermN other non)
economic variable that is considered is (nde)* : %allup rating of 506 is considered a safe floor for the
incumbentAs popularity ,(e"is)*ec& T+ice, 1-.2/ and as observed from the historical %allup values, .06
10

is a conservative cap on the 9ob approval rating So, "hile defining the (nde) variable, the lo"er cutoff is
706 rating and the higher cutoff is E06 rating
1ence, the variable (nde) ta&es indicator values bet"een 0 and 2 as follo"sJ
For A:crogc_0ollup 4uJ ;nde$ O 0
4u < A:crogc_0ollup < 6uJ ;nde$ O 1
A:crogc_0ollup 6uJ ;nde$ O 2
The follo"ing models are considered to e$plore the influence of non)economic variables on vote
percentageJ
Table *J :nalysis of ;nfluence of Gon)#conomic 2ariables
Model Pear +
2
,6/ p)Statistic
I0IE =
1
+
2
0AIIuP_[uNE +
3
wARS 1-7.)200. EE28.
%:((U!
Q

3:+S
00002
0527
I0IE =
1
+
2
0AIIuP_[uNE +
3
SCANAIS 1-7.)200. <EEE5
%:((U!
Q

SB:GD:(S
Q

0000.
0025
I0IE =
1
+
2
0AIIuP_[uNE +
3
INEX 1-7.)200. <7<<5
%:((U!
Q

;GD#S
Q

00001
0077
I0IE =
1
+
2
0AIIuP_[uNE +
3
INEX
+
4
SCANAIS
1-7.)200. .8E88
%:((U!
Q

;GD#S
Q

SB:GD:(S
Q

00002
0078
0025
I0IE =
1
+
2
AIERA0E_0AIIuP +
3
INEX 1-7.)200. E212E
:2#+:%#R%:((U!
Q

;GD#S
Q

0000-

000<-
I0IE =
1
+
2
HIIERH +
3
SCANAIS 1-7.)200. 5E---
M;DT#+M
QQ

SB:GD:(S
Q

005E
0002
*Denotes signifiant !"#a$%e at & ' $e#e$ of signifiane
** Denotes signifiant !"#a$%e at + ' $e#e$ of signifiane
:s observedN 0une %allup rating, average %allup rating in the term of the incumbency, scandals and
midterm are significant variables The follo"ing table summari=es the correlations bet"een the
significant economic and non)economic variablesJ
%allupR0une :verageR%allup ;nde$ Scandals Midterm %ro"th
%allupR0une 1000
:verageR%allup 0.0E
QQ
1000
;nde$ 0EE< 0-11
QQ
1000
Scandals )08-8 )0871 )0150 1000
Midterm )01.0 )00E5 )00.. )021. 1000
%ro"th 0287 01.2 0028 )01-5 )00E2 1000
** ,orre$ation more than -*.
11

:s observed, the correlation bet"een :verageR%allup T %allupR0une and bet"een :verageR%allup T
;nde$ is above the acceptable cutoff of 0. 1ence, only one of the %allup ratings could be included in
the forecasting model to avoid multicollinearity Since, the %allup rating is a more relevant factor prior
to the election than the :verage)%allup ratingN it is preferred in the forecasting model
$. Proposed Model
The driving criterion for a robust forecasting model is a high value of +
2
, significant p)values of the
coefficients of the independent variables, acceptable levels of +oot Mean S>uare #rror ,+MS#/, lo"er
Theil statistic and a logical relationship bet"een the dependent and independent variables
*ased on the analysis and results in Section 8, the follo"ing model is proposed for forecasting the US
!residential elections that combine the significant economic and non)economic factorsJ
I0IE =
1
+
2
0R0wIE +
2
0AIIuP +
3
SCANAIS +
4
HIIERH+
5
INEX +ERR0R
The %D! gro"th rate is an indicator of the health of the economy and the 0une %allup 9ob approval
rating is a reflection of the popularity and performance of the incumbent !resident prior to the
forthcoming election The scandals affect the chances of the incumbent party for re)election and by an
even bigger degree in case the running candidate is the incumbent !resident Midterm gives an
indication of the performance of the t"o parties during the last nation)"ide election and is similar to a
referendum on the partyAs performances ;nde$ is also included as an independent variable The variable
combines the %allup rating during 0une of the election years "ith the average rating during the tenure
This accounts for ma9or policy decisions that "ere ta&en during the initial >uarters of the term
For the period 1-7.)200., the model e$hibits a +
2
of -7<-7 6 and ad9usted +
2
of -21-2 6 :ll
independent variables in the above model are highly significant at 5 6 level of significance e$cept
M;DT#+M "hich is significant at <6 level of significance The Durbin 3atson statistic of the model is
227- and model F statistic is 8E720 "ith p)value 0000007 This sho"s that model is highly significant
and can be used for forecasting US !residential elections The above regression results for the model
are summari=ed in Table B


12

Table BJ !roposed #stimated Model using Data from 1-7.)200. for Forecasting 2012 #lection
Dependent 2ariableJ 24T# ;ncluded observationsJ 1E
MethodJ (east S>uares
2ariable Boefficient Std #rror t)Statistic !rob
B 70E.< 215- 1..7. 0000
%+43T1 0E7E 01E2 8-.1 0008
%:((U!0UG# 0808 007. E8.. 0000
SB:GD:( )8110 0<58 )712. 0002
%;GD#S )2202 0.15 )2<02 0022
M;DT#+M )1000 07.2 )20<7 00E5
+)s>uared 0-7<-7 Mean dependent var 520-1E-0
:d9usted +)s>uared 0-21-2 SD dependent var 55-1E-E
S# of regression 15E252 :&ai&e info criterion 70107<.
Sum s>uared resid 27717.0 Sch"ar= criterion 78001--
(og li&elihood )2E0.8.2 F)statistic 8E71-.10
Durbin)3atson stat 227-00 !rob,F)statistic/ 0000007

200% Presidential Election

The 200. election "as a closely fought election bet"een the Democratic candidate Mr *arac& 4bama
and +epublican candidate Mr 0ohn McBain The independent variables for the election of 200. ta&e the
follo"ing valuesJ
;ndependent 2ariable 2alue
%ro"th ,6/ 022
%allup 2.000
Scandals 1
Midterm )0.1E
;nde$ 0

Using the data from 1-7.)2007, the model developed predicts 7<.806 vote for the incumbent party
and a victory for Mr *arac& 4bama The forecasting model has the follo"ing statisticsJ
+oot Mean S>uare #rrorJ 1250 6
Mean :bsolute #rrorJ 1025 6
Theil ;ne>uality BoefficientJ 0012

18

"ith the follo"ing model parametersJ
Table DJ !roposed #stimated Model using Data from 1-7.)2007 for Forecasting 200. #lection
Dependent 2ariableJ 24T# ;ncluded observationsJ 15
MethodJ (east S>uares
2ariable Boefficient Std #rror t)Statistic !rob
B 7185. 250E 1E507 0000
%+43T1 0E20 01<8 85<2 000E
%:((U!0UG# 02-. 0050 5-<5 0000
%;GD#S )28.7 0.-< )2E5E 002E
M;DT#+M )10-- 052E )20.- 00EE
SB:GD:( )8217 0<-- )7028 0008
+)s>uared 0-7E1- Mean dependent var 5275<.0
:d9usted +)s>uared 0-1E80 SD dependent var 55.5-8
S# of regression 1E1E0- :&ai&e info criterion 70.<0<
Sum s>uared resid 28505<< Sch"ar= criterion 78<02-
(og li&elihood )27E5805 F)statistic 81E51E8
Durbin)3atson stat 270701 !rob,F)statistic/ 000002

The actual results "ere slightly different from this predicted value and 0ohn McBain lost to *arac&
4bama "ith the vote percentage being 7EE)587 The developed model predicts the 200. election
closely ;t combines the significant economic and non)economic variables and offers a more holistic
forecasting model than those present in the contemporary literature The above results validate the
proposed model 1ence, it can be used for forecasting the 2012 US !residential election
&. Forecasting 2012 Presidential Election
The 2012 election is being contested bet"een the Democratic candidate Mr *arac& 4bama and
+epublican candidate Mr Mitt +omney The incumbent, !resident 4bama is see&ing reelection after
holding office during a period of slo" economic recovery Mr +omney is a successful businessman
turned politician and is a former %overnor of Massachusetts



17

The values of the independent variables in the proposed model for the election of 2012 are as follo"sJ
;ndependent 2ariable 2alue
%ro"th ,6/ 1E2
%allup 7.000
Scandals 0
Midterm )0E82
;nde$ 1
The proposed model forecasts that the vote percentage share of the incumbent Democratic !arty
candidate Mr *arac& 4bama in the t"o)party !residential election for 2012 is li&ely to be 5728-6 The
forecast statistics given by the proposed model are as follo"sJ
+oot Mean S>uare #rror ,+MS#/J 12856
Mean :bsolute #rrorJ 0--26
Theil ;ne>uality BoefficientJ 0011
The model parameters are presented in Table B above
The forecast of vote share of the incumbent Democratic candidate is 5728- 6, "ith -56 lo"er
confidence interval on forecast to be 51.1. 6 )5728- 6 1ence, the model predicts the victory of the
incumbent !resident Mr *arac& 4bama
onclusion
The model predicts a comfortable victory for the Democrat party candidate Mr *arac& 4bama in the
2012 election The proposed model also predicted the 200. !residential election successfully, "ith the
predicted incumbent vote percentage being 7<.80 6, that is close to the actual vote percentage share
,7EE 6/ received by then +epublican party candidate Mr 0ohn McBain
The model suggested illustrates the follo"ing features for US !residential election outcome forecastsJ
a/ US !residential election results are not 9ust decided by economic variablesJ 4ur study sho"s
that %D! gro"th rate is the only &ey economic factor 2arious other economic factors such as
unemployment rate, interest rate, public debt, budget deficit'surplus, e$change rate, inflation,
percentage change in oil price and gold price, healthcare spending "ere insignificant This is in
contrast "ith the model presented by Fair ,1-<./ and the contemporary discussions about the
15

forthcoming 2012 election that suggest that the US !residential election results are mostly
determined by economic conditions li&e unemployment rate and inflation
b/ ;mpact of non)economic variablesJ The model sho"s that non)economic factors play a large part
in determining election results The important factors are not only the !residential scorecard
determined by the %allup 9ob approval rating in 0une ,as suggested by (e"is)*ec& et al ,1-.2//,
but also non)economic factors li&e average rating during the tenure, presence'absence of
scandals lin&ed to the incumbent !resident and the midterm performance of the parties
These results signal a shift in the US !residential forecasting research since the 200. !residential
election The emergence of non)economic factors highlights the changing dynamics of US
!residential election outcomes
References
:brahamo"it=, :lan ; ,1-../ :n ;mproved Model for !redicting the 4utcomes of !residential
#lections PS/ Po$itia$ Siene and Po$itis, 01 2, .78).7<
*an& of #ngland ,2010/ The U3 reession in onte)t 4 5hat do three ent%ries of data te$$ %s6
retrieved from
httpJ''"""ban&ofenglandcou&'publications'Documents'>uarterlybulletin'threecenturiesofd
ata$ls
*artels, ( M T Haller, 0 ,2001/ !residential 2ote ModelsJ : +ecount PS/ Po$itia$ Siene and
Po$itis, 777(V 819, -U28
*ureau of (abor Statistics ,2012a/ Ho5 the :o#ernment ;eas%res Unem!$o<ment* retrieved
from httpJ''"""blsgov'cps'cpsRhtgmhtmVunemployed
*ureau of (abor Statistics ,2012b/ =here an ( find the %nem!$o<ment rate for !re#io%s <ears6*
retrieved from httpJ''"""blsgov'cps'prevRyrshtm'
*ureau of #conomic :nalysis ,2012/ Ta>$e ?*10* :o#ernment Soia$ @enefits* retrieved from
httpJ''"""beagov'national'inde$htmVgdp
Bampbell, 0 # ,1--2/ Forecasting the !residential 2ote in the States Amerian Ao%rna$ of
Po$itia$ Siene, ?+ 0,8.E)70<
Bu=Wn, : %, 1eggen +0, T *undric&,BM ,2000/ Fiscal policy, economic conditions, and terms
in officeJ simulating presidential election outcomes* (n Proeedings of the =or$d ,ongress of the
S<stems Sienes and (SSS (nternationa$ Soiet< for the S<stems Sienes, 77th :nnual Meeting,
0uly 1EU20, Toronto, Banada
1E

#ri&son, + S, and 3le=ien, B ,1--E/ 4f time and presidential election forecasts PS/ Po$itia$
Siene and !o$itis, ?1, 8<)8-
Fair, + B ,1-<./ The effect of economic events on votes for president Re#ie5 of Eonomis
and Statistis, +-, 15-)1<8
Fair, + B ,2002/ !redicting !residential #lections and 4ther Things Stanford/ Stanford
Uni#ersit< Press*
Fair, +B ,200E/ The Effet of Eonomi E#ents on Votes for President/ 0--2 U!date retrieved
from J httpJ''fairmodeleconyaleedu'+:PF:;+'!DF'200EB1TM1TM
Fair, +B ,200./ 0--B Post ;ortem retrieved from J
httpJ''fairmodeleconyaleedu'vote200.'inde$2htm
Fair, + B ,2012/ Vote"Share EC%ations/ No#em>er 0-1- U!date* retrieved from
httpJ''fairmodeleconyaleedu'vote2012'inde$2htm
Federal +eserve ,2012/ Historia$ Data* retrieved from
httpJ''"""federalreservegov'releases'h15'datahtm
%allup !residential !oll ,2012/ Presidentia$ Ao> A!!ro#a$ ,enter retrieved from
httpJ''"""gallupcom'poll'127-22'presidential)approval)centerasp$
1ibbs, Douglas : ,2000/ *read and !eace voting in US presidential elections P%>$i ,hoie,
1-2, 17-U1.0
1ibbs, Douglas : ,2012/ 4bamaAs +e)election !rospects Under I*read and !eaceA 2oting in the
2012 US !residential #lection retrieved fromJ httpJ''"""douglas)
hibbscom'1ibbs:rticles'1;**SR4*:M:)+##(#BT)810uly2012r1pdf
;nternational Monetary Fund ,2010/ Historia$ P%>$i De>t Data>ase retrieved from
httpJ''"""imforg'e$ternal'pubs'ft'"p'2010'data'"p10275=ip
;nflationDatacom ,2012/ Historia$ ,r%de Oi$ Pries 8Ta>$e9* retrieved from
httpJ''inflationdatacom'inflation';nflationR+ate'1istoricalR4ilR!ricesRTableasp
0CrDme, *runo T 0CrDme )Spe=iari, 2Croni>ue,2011/ Forecasting the 2012 US !residential
#lectionJ 3hat Ban 3e (earn from a State (evel !olitical #conomy Model (n Proeedings of the
APSA Ann%a$ meeting Seatt$e, September 1)7 2011
Feilis)*oro&, 2 ; T (ichtman, : 0 ,1-.1/ !attern +ecognition :pplied to !residential #lections
in the United States, 1.E0)1-.0J The +ole of ;ntegral Social, #conomic, and !olitical Traits
Proeedings of the Nationa$ Aadem< of Sienes, <., <280X<287
(e"is)*ec&, M S T +ice, T 3 ,1-.2/!residential !opularity and !residential 2ote The P%>$i
O!inion D%arter$<, 2+ 2, 587)58<
1<

(ichtman, : 0 ,2005/ The Feys to the 3hite 1ouse Lanham, ;D/ Le)ington @ooks
(ichtman, : 0 ,200./ The &eys to the "hite houseJ :n inde$ forecast for 200. (nternationa$
Ao%rna$ of Eoreasting, 02, 801U80-
4ffice of the Bler& ,2010/ E$etion Statistis retrieved from
httpJ''artandhistoryhousegov'houseRhistory'election;nfo'inde$asp$
Sigelman, (, ,1-<-/ !residential popularity and presidential elections P%>$i O!inion D%arter$<,
2?, 582)87
Silver, G ,2011/ On the ;addening$< (ne)at Re$ationshi! @et5een Unem!$o<ment and Re"
E$etion* retrieved fromJ httpJ''fivethirtyeightblogsnytimescom'2011'0E'02'on)the)
maddeningly)ine$act)relationship)bet"een)unemployment)and)re)election'
Sinha, ! and *ansal,:F ,200./ 1ierarchical *ayes !rediction for the 200. US !residential
#lection The Ao%rna$ of Predition ;arkets, 0, 7<)E0
The 3hite 1ouse ,2012/ Ta>$e 1*04S%mmar< of Reei!ts, O%t$a<s, And S%r!$%ses Or Defiits 8F9
As Perentages Of :DP/ 1G?-F0-1. retrieved from
httpJ''""""hitehousegov'sites'default'files'omb'budget'fy2018'assets'histpdf
Tufte, # + ,1-<5/ Determinants of the 4utcomes of Midterm Bongressional #lections
Amerian Po$itia$ Siene Re#ie5, E-, .12)2E
United States Gational Mining :ssociation ,2011/ Historia$ :o$d Pries" 1B?? to Present*
retrieved from httpJ''"""nmaorg'pdf'gold'hisRgoldRpricespdf
1.

Appendi'
(able 1) Scandals during Presidential (er*s and the orresponding Ratings
#lection
Pear
;ncumbent !resident Scandals
Scandal
+ating
1-7.
Fran&lin D +oosevelt
*udget cuts for the military
+ecognition of ;srael
Taft) 1arley :ctJ +educing the po"er of the labor
unions
1
1arry S Truman Gone
1-52 1arry S Truman
Bontinuous accusations of spies in the US %ovt
Foreign policiesJ Forean "ar, ;ndo Bhina "ar
3hite house renovations
Steel and coal stri&es
Borruption charges
1
1-5E D"ight D #isenho"er Gone 0
1-E0 D"ight D #isenho"er
U)2 Spy !lane ;ncident
Senator 0oseph + McBarthy Bontroversy
(ittle +oc& School +acial ;ssues
1
1-E7
0ohn F Fennedy #$tra marital relationships
0
(yndon * 0ohnson Gone
1-E. (yndon * 0ohnson
2ietnam "ar
Urban riots
!hone Tapping
1
1-<2 +ichard Gi$on Gi$on shoc& 0
1-<E
+ichard Gi$on 3atergate Scandal
2
%erald Ford Gi$on !ardon
1-.0 0immy Barter
;ran hostage crisis
1-<- energy crisis
*oycott of the Mosco" 4lympics
1
1-.7 +onald +eagan
Ta$ cuts and budget proposals to e$pand military
spending
0
1-.. +onald +eagan
;ran)Bontra affair
Multiple corruption charges against high ran&ing
officials
1
1--2 %eorge 1 3 *ush
+enegation on election promise of no ne" ta$es
Y2omiting ;ncidentY
1
1--E *ill Blinton
Firing of 3hite 1ouse staff
YDonZt as&, donZt tellY policy
1
2000 *ill Blinton (e"ins&y Scandal 2
2007 %eorge 3 *ush !oor handling of Fatrina 1urricane) Gone 0
200. %eorge 3 *ush
Midterm dismissal of < US attorneys
%uantanamo *ay Bontroversy and torture
1
2012 *arac& 4bama Gone 0
1-


(able 2) Militar# Inter+entions during Presidential (er*s and the
orresponding Ratings
#lection
Pear
;ncumbent !resident Military ;nterventions
3ar
+ating
1-7.
Fran&lin D +oosevelt 3orld 3ar 2
1
1arry S Truman 1iroshima'Gagasa&i
1-52 1arry S Truman Forean 3ar )1
1-5E D"ight D #isenho"er #nded Forean 3ar 1
1-E0 D"ight D #isenho"er Gone 0
1-E7
0ohn F Fennedy
*ay of !igs
Buban Missile crisis
2ietnam
)1
(yndon * 0ohnson 2ietnam
1-E. (yndon * 0ohnson
2ietnam
;sarel
)1
1-<2 +ichard Gi$on 2ietnam )1
1-<E
+ichard Gi$on 2ietnam
1
%erald Ford 2ietnam ,end/
1-.0 0immy Barter Gone 0
1-.7 +onald +eagan Bold 3ar 0
1-.. +onald +eagan Bold 3ar 0
1--2 %eorge 1 3 *ush
!anama
%ulf 3ar
Somalia
)1
1--E *ill Blinton
Somalia
*osnia
0
2000 *ill Blinton Serbians ,Pugoslavia/ 0
2007 %eorge 3 *ush
:fghanistan
;ra>
1
200. %eorge 3 *ush
:fghanistan
;ra>
)1
2012 *arac& 4bama
#nded ;ra> "ar
;ncreased presence in :fghanistan
Military ;ntervention in (ibya
1



20

(able ") ,allup Ratings
#lection
Pear
;ncumbent !resident
!eriod of %allup
Measurement
+ating
0une %allup +ating
:verage %allup
+ating
%allup ;nde$
1-7. 1arry S Truman May 2<)0une1
8-
8-5 55E 1
0une 1<)28 70
1-52 1arry S Truman
May 2-)0une 8 81
815 8E5 0
0une 17)1- 82
1-5E D"ight D #isenho"er
May 80)0une 7 <1
<2 E-E 2
0une 17)1- <8
1-E0 D"ight D #isenho"er
0une 15)20 E1
5- E05 2
0une 2-)0uly 7 5<
1-E7 (yndon * 0ohnson
0une 8). <7
<7 <72 2 0une 10)15 <7
0une 27)2- <7
1-E. (yndon * 0ohnson
0une 12)1< 72
71 508 1
0une 25)80 70
1-<2 +ichard Gi$on
0une 15)1. 5-
5<5 55. 1
0une 22)25 5E
1-<E %erald Ford 0une 10)18 75 75 7<2 1
1-.0 0immy Barter
May 2-)0une 1 8.
88E 755 1 0une 12)15 82
0une 2E)2- 81
1-.7 +onald +eagan
0une 5)< 55
57 508 1 0une 21)27 57
0une 2.)0uly 1 58
1-.. +onald +eagan
0une -)12 51
50 558 1 0une 28)2E 7.
0une 80)0ul E 51
1--2 %eorge 1 3 *ush
0une 8)E 8<
8<8 E0- 2 0une 11)18 8<
0une 25)2- 8.
1--E *ill Blinton
0une 1<)1. 5.
55 7-E 1
0une 2E)2- 52
2000 *ill Blinton
0une 5)E E0
5<5 E0E 2
0une 21)27 55
2007 %eorge 3 *ush
0une 2)5 7-
7.5 E22 2
0une 20)22 7.
200. %eorge 3 *ush
0une .)11 80
2- 8E5 0
0une 17)1. 2.
2012 *arac& 4bama
May 2<)0une 2 7E
7E7 7-0 1
0une 8)- 7<
0une 10)1E 7E
0une 1<)28 7E
0une 27)80 7<
SourceJ %allup !residential !oll ,2012/

21

(able $) Midter* Elections Results -1.$$/20100
Pear
;ncumbent
!arty
Midterm
#lection Pear
1ouse Seats 1ouse
+esult
Senate Seats Senate
+esult
Midterm
2alues Democrati
c
+epublican Democ
ratic
+epu
blican
1-7. Democratic
1-77 278 1-0
)1
5< 8.
)1 )100
1-7E 1.. 27E 75 51
1-52 Democratic
1-7. 2E8 1<1
1
57 72
1 100
1-50 287 1-- 7. 7<
1-5E +epublican
1-52 218 221
)1
7E 7.
)1 )100
1-57 282 208 7. 7<
1-E0 +epublican
1-5E 287 201
)1
7- 7<
)1 )1
1-5. 2.8 158 E7 87
1-E7 Democrat
1-E0 2E2 1<5
1
E7 8E
1 100
1-E2 25. 1<E E< 88
1-E. Democrat
1-E7 2-5 170
1
E. 82
1 100
1-EE 27. 1.< E7 8E
1-<2 +epublican
1-E. 278 1-2
)1
5. 72
)1 )100
1-<0 255 1.0 57 77
1-<E +epublican
1-<2 272 1-2
)1
5E 72
)1 )100
1-<7 2-1 177 E1 8<
1-.0 Democrat
1-<E 2-2 178
1
E1 8.
1 100
1-<. 2<< 15. 5. 71
1-.7 +epublican
1-.0 272 1-2
)1
7E 58
1 )0E8
1-.2 2E- 1EE 7E 57
1-.. +epublican
1-.7 258 1.2
)1
7< 58
)1 )0E8
1-.E 25. 1<< 55 75
1--2 +epublican
1-.. 2E0 1<5
)1
55 75
)1 )100
1--0 2E< 1E< 5E 77
1--E Democrat
1--2 25. 1<E
)1
5< 78
)1 )100
1--7 207 280 7. 52
2000 Democrat
1--E 20< 22E
)1
75 55
)1 )100
1--. 211 228 75 55
2007 +epublican
2000 212 221
1
50 50
1 100
2002 207 22- 7. 51
200. +epublican
2007 202 282
)1
77 55
0 )0.2
200E 288 202 7- 7-
2012 Democrat
200. 25E 1<.
)1
55 71
1 )0E8
2010 1-8 272 51 7<
SourceJ 4ffice of the Bler& ,2010/





22

(able &a) Macroecono*ic 1ariables
Pear
Unemployment
+ate ,6/
a

;nterest
+ate ,6/
b

;nflation
,6/
c

1ealthcare
*udget ,6/
d

%ro"th +ate
,6/
e

1-77 G': G': 0000 122. 72<-
1-7. 8.00 G': 0000 8E<- 85<-
1-52 8000 G': 28E2 80<0 0E-1
1-5E 7100 2<80 1-85 85E< )1751
1-E0 5500 8210 1-E< 7E-2 08<<
1-E7 5200 8500 12E0 7<<< 510-
1-E. 8E00 5EE0 818- 5-27 5078
1-<2 5E00 7770 7.15 <<8- 5-17
1-<E <<00 5050 <E80 -.57 8<51
1-.0 <100 18850 <.81 -.72 )85-<
1-.7 <500 10280 525- -<<1 5770
1-.. 5500 <5<0 2-0E -71- 21<.
1--2 <500 8520 82.0 11E00 2EE2
1--E 5700 5800 20E2 115-7 8121
2000 7000 E270 1E05 10578 121-
2007 5500 1850 2825 11..- 2E-0
200. 5.00 1-20 8052 18000 0220
2012 G': G': G': G': 1E20
aJ *ureau of (abor Statistics ,2012b/, bJ Federal +eserve ,2012/, cJ Fair,200E,200./, dJ *ureau of
#conomic :nalysis ,2012/, eJ Fair ,200E, 200.,2012/












28

(able &b) Macroecono*ic 1ariables
Pear
2ote ,6 share
of incumbent
party/
a

*udget
Surplus'Defici
t ,6/
b

!ublic
Debt ,6/
c

%old
!rices ,K
per
4unce/
d

4il !rices
,K'bbl/
e

#$change
+ate ,K'L/
f

1-77 58<<7 )22<00 -17-0 88.50 G': 7082
1-7. 528<0 7E00 -85.0 87<10 2<<0 7082
1-52 775-5 )0700 <2255 87E00 2<<0 2<-8
1-5E 5<<E7 0-00 E22<2 87--0 2-70 2<-8
1-E0 7--18 0100 572-1 852<0 2-10 2.0-
1-E7 E1877 )0-00 7E-1E 85100 8000 2<-8
1-E. 7-5-E )2-00 8.188 8-810 81.0 28-2
1-<2 E1<.- )2000 85175 5.720 8E00 2500
1-<E 7.-7. )7200 877.5 127<70 18100 1.05
1-.0 77E-< )2<00 722<< E15000 8<720 282E
1-.7 5-1<0 )7.00 50.-E 8E1000 2.<50 188<
1-.. 58-02 )8100 E1-71 78<000 17.<0 1<.8
1--2 7E575 )7<00 <0<8E 878.20 1-250 1<E<
1--E 57<8E )1700 <02-- 8.<.10 207E0 15E8
2000 502E5 2700 57.85 2<-110 2<8-0 1515
2007 51288 )8500 E1720 70-<20 8<EE0 1.82
200. 7EE00 )8200 <1221 .<1-E0 -17.0 1.52
2012 G': G': G': G': G': G':
aJ Fair ,200E, 200./, bJ The 3hite 1ouse ,2012/, cJ ;nternational Monetary Fund ,2010/, dJ United States
Gational Mining :ssociation,2011/,eJ ;nflationDatacom,2012/, fJ *an& of #ngland,2010/

Vous aimerez peut-être aussi