Tải bản đầy đủ - 0 (trang)
2 Data: The American Panel Survey (TAPS)

2 Data: The American Panel Survey (TAPS)

Tải bản đầy đủ - 0trang


J.H. Kim and N. Schofield

Fig. 1 Voter distribution and candidate positions in 2012 U.S. presidential election

The positions of the major presidential candidates, Romney and Obama, in 2012

are estimated by taking the means of policy positions of respondents who voted for

each candidate. These positions are denoted (z*Obama , z*Romney ), and the candidate

position matrix is given by:


Candidate Obama Romney

z D 4 x W Economic 0:48 0:71 5

y W Social





From the matrix, we see that Obama’s estimated position is more liberal than

Romney’s position in terms of both social and economic policy area, as expected.

Next, we estimate a multinomial logit model (MNL) to examine the effect of

ideological position and valence of candidates on vote choice. Table 1 summarizes

the results. From the first column in Table 1, we see that the spatial coefficient is

1.117 and statistically significant. Here, the reference category is Vote for Obama.

The relative valence of Romney is 0.044. Compared to the 2008 presidential

election (Schofield and Gallego 2011), the magnitude of spatial coefficient is

greater in 2012 election. However, unlike 2008 election, the valence term in 2012

election is not statistically significant. The spatial coefficient remains statistically

significant when we condition demographic variables, such as ethnicity, education,

age, and gender in the model. From the second column in Table 1, we see that

black respondents are less likely to vote for Romney, and respondents with college

education are more likely to vote for Romney than Obama.

Spatial Model of U.S. Presidential Election in 2012


Table 1 Spatial model of 2012 U.S. presidential election


Romney valence

Spatial ˇ

Obama trait

Romney trait

Ethnicity (black)

Ethnicity (Hispanic)

Ethnicity (white)

Ethnicity (other)

Education (less than

high school)

Education (high


Education (college)


Gender (female)


Log likelihood

McFadden R2

(1) Pure spatial

0.044 (0.117)

1.117*** (0.075)

(2) Spatial C traits

0.422* (0.192)

0.575*** (0.116)

3.237*** (0.354)

2.844*** (0.321)

(3) Spatial C traits C socios

2.899 (5.757)

0.721*** (0.156)

3.640*** (0.481)

3.070*** (0.413)

2.503 (5.809)

0.878 (5.669)

3.287 (5.641)

2.623 (5.699)

4.939*** (1.451)

0.689 (0.702)







0.555 (0.572)

0.013 (0.014)

0.606 (0.466)




Note: Baseline category is Vote for Obama, Standard error in parentheses

*p < 0.05; **p < 0.01; ***p < 0.001

Also, from the results, we can calculate the probability that a voter chooses each

candidate, when all candidates are located at the electoral mean of policy positions.

First, the probability that a voter chooses Obama is




D 0:511

exp.0/ C exp . 0:044/


Using the same calculation, the probability that a voter chooses Romney is



exp . 0:044/

D 0:489

1 C exp . 0:044/


The results of calculation are close to the actual result from the sample. While

56 % of the respondents answered they voted for Barack Obama, 44 % of them

answered that they voted for Mitt Romney at 2012 presidential election.


J.H. Kim and N. Schofield

In order to check whether the joint origin, z0 is a Local Nash Equilibrium (LNE),

we examine the Hessians of the vote share utility functions. The Hessian, or the

characteristic matrix of Romney’s vote share function at z0 is given by:

r I


0:752 0:128

.1 2 0:489/

0:128 0:751


0:963 0:006


0:006 0:964

cRomney D 2ˇ 1






The eigenvalues of the characteristic matrix are 0.957 with the eigenvector

( 0.747, 0.665), and 0.970 with the eigenvector (0.665, 0.747). According to

Schofield (2007), negative eigenvalues of the characteristic matrix is a necessary

condition for the electoral mean to be SLNE.

The convergence coefficient can be calculated by using the following formula:


c D 2ˇ 1 2 Romney Tr Œr

1:117 .1 2 0:489/ 1:473 D 0:073


Schofield et al. (2011a, b) show that c < 1 is a sufficient condition for convergence

to z0 in the pure spatial model. Using simulations, we find a LNE with the following

candidate positions:



Candidate Obama Romney

z1 D 4 Economic W 0:032 0:032 5

Social W 0:004 0:004


As we see, the candidate positions in this LNE are very close to the joint electoral


We extend the spatial model by including perceptions of candidate traits. Using

the respondents’ evaluations of candidates, we constructed a trait index by factor

analysis (see the appendix for factor loadings). From Table 1, we find that the spatial

coefficient is still significant in models with candidate trait variables.

4 Discussion

From the analysis of voter perceptions in 2012 U.S. presidential election, we can

draw the following preliminary findings. First, as in the case of previous presidential

elections, ideological distance between candidate and voters plays a significant role

in determining vote choice in 2012 election. The magnitude of spatial coefficient

turns out to be even greater than the previous election. Second, the valence of

candidate in 2012 election is not a statically significant predictor of vote choice.

Spatial Model of U.S. Presidential Election in 2012


These findings provide implications about the influence of increased campaign

contribution, resulting from Supreme Court decision in 2011. That is, the exogenous

increase in campaign contribution has emphasized the role of ideological distance

in voting behavior, and decreased the effect of valence.

Appendix 1: Questions for the 2012 TAPS

1. Indicate your level of agreement with this statement: Federal personal income

taxes for individuals with incomes higher than $250,000 should be raised.

2. Which actions are you in favor of and which are you against: less government

regulation of business?

3. Indicate your level of agreement: Incomes should be more equal because

everyone’s contribution to society is equally important.

4. Do you consider your view of the federal government recognizing same-sex

marriages liberal, moderate, or conservative?

5. Do you consider your view of federal funding for abortion liberal, moderate, or


6. Do you consider your view of the banning possession of handguns liberal,

moderate, or conservative?

7. Do you consider your view of programs designed to help minorities get better

jobs and education liberal, moderate, or conservative?

Appendix 2: Factor Loadings for Economic and Social Policy

See Tables 2 and 3.

Table 2 Factor loadings for economic and social policy


Personal income tax

Government regulation of business

Income inequality

Gay marriage


Gun control

Helping minorities

Economic policy








Social policy









Table 3 Factor loadings for

candidate traits

J.H. Kim and N. Schofield























Obama trait





















Romney trait






















Abramowitz, A. I. (1991). Incumbency, campaign spending, and the decline of competition in US

House elections. Journal of Politics, 53(1), 34–56.

Gerber, A. (1998). Estimating the effect of campaign spending on senate election outcomes using

instrumental variables. American Political Science Review, 92(2), 401–411.

Hansen, W. L., Rocca, M. S., & Ortiz, B. L. (2015). The effects of Citizens United on corporate

spending in the 2012 presidential election. Journal of Politics, 77(2), 535–545.

Jacobson, G. C. (1990). The effects of campaign spending in House elections: New evidence for

old arguments. American Journal of Political Science, 34, 334–362.

Klumpp, T., Mialon, H. M., & Williams, M. A., et al. (2012). Money talks: The impact of Citizens

United on state elections. Emory University. Journal of Law and Economics, forthcoming.

Krasno, J. S., & Green, D. P. (1988). Preempting quality challengers in House elections. Journal

of Politics, 50(4), 920–936.

La Raja, R. J., & Schaffner, B. F. (2014). The effects of campaign finance spending bans on

electoral outcomes: Evidence from the states about the potential impact of Citizens United

v. FEC. Electoral Studies, 33, 102–114.

Miller, G., & Schofield, N. (2003). Activists and partisan realignment in the United States.

American Political Science Review, 97(2), 245–260.

Schofield, N. (2007). The mean voter theorem: Necessary and sufficient conditions for convergent

equilibrium. Review of Economic Studies, 74(3), 965–980.

Schofield, N., & Gallego, M. (2011). Leadership or Chaos. Berlin: Springer.

Schofield, N., Claassen, C., Gallego, M., & Ozdemir, U. (2011a). Empirical and formal models

of the US presidential elections in 2000 and 2004. In N. Schofield & G. Caballero (Eds.), The

political economy of institutions, democracy and voting (pp. 217–258). Berlin: Springer.

Spatial Model of U.S. Presidential Election in 2012


Schofield, N., Claasen, C., Ozdemir, U., & Zakharov, A. (2011b). Estimating the effects of activists

in two-party and multi-party systems: A comparison of the United States in 2008 and Israel in

1996. Social Choice and Welfare, 36(3–4), 483–518.

Schofield, N., & Miller, G. (2007). Elections and activist coalitions in the United States. American

Journal of Political Science, 51(3), 518–531.

Modeling Elections and Referenda in Ireland

Norman Schofield and William Simoneau

1 Introduction

The period between 2007 and 2009 was a tumultuous time for the Republic of

Ireland. During this time the Irish economy was fully exposed to the global financial

crisis. In turn, Irish households saw an overall drop in net worth of e453 billion

between 2007 and 2010 (Cussen and Phelan 2010). In January of 2008 the Irish

unemployment rate started a steady climb from 4 % up to 15.1 % in 2012 (CSO

Data). The financial crisis marked the end of the Celtic Tiger, a period of economic

growth and modernization that had begun in the early 1990s. During this sharp

negative downturn the Irish electorate went to the polls on three separate occasions.

In 2007, there was the general election for the lower house, the Dáil Éireann, or

Dáil, in 2008, a referendum on the ratification of the Lisbon Treaty and in 2009 a

second referendum on the Lisbon Treaty.

Signed by EU member states in December 2007, the Lisbon Treaty, proposed

changes to European Union institutions and attempted to streamline the internal

workings of the EU after enlargement in 2004 and 2007 (Quinlan 2009). The Lisbon

Treaty was the result of the redrafting of the failed attempt at a Constitutional Treaty.

The Constitutional Treaty, agreed upon in 2004, was the first attempt to streamline

the decision making process in the wake of the 2004 EU enlargement that added ten

new member states. The Constitutional Treaty failed in the ratification process when

Dutch and French voters voted down the Treaty in a national referendum. In the

wake of the failed ratification process, the Lisbon Treaty was developed and passed.

Although there were differences between the two treaties their goals of streamlining

the institutions remained the same. Although these differences were disputed,

N. Schofield ( ) • W. Simoneau

Washington University in Saint Louis, Campus Box 1063, One Brookings Drive, St. Louis,

MO 63130-4899, USA

e-mail: schofield.norman@gmail.com

© Springer International Publishing Switzerland 2016

M. Gallego, N. Schofield (eds.), The Political Economy of Social Choices,

Studies in Political Economy, DOI 10.1007/978-3-319-40118-8_11



N. Schofield and W. Simoneau

former French President Valery Giscard d’Estaing said, “The Treaty of Lisbon is

the same as the rejected constitution. Only the format has been changed to avoid

referendums” (Bonde 2010). Further, Alexander Stubb, former Finnish minister of

foreign affairs, stated that 99 % of the Constitution had been kept in the Lisbon

Treaty (Bonde 2010). One key difference was the ratification process. The Lisbon

Treaty was passed by all member states, except Ireland, through Parliamentary votes

instead of national referenda. The Irish Constitution necessitated a referendum in

order to pass the treaty.

According to Article 46 of the Irish Constitution any provision of the Constitution

may be amended, whether by way of variation, addition, or repeal. It also provides

that every proposal for amendment must be initiated in the Dáil as a Bill. Having

been passed or deemed to have been passed by both houses of the Oireachtas, the

Bill is submitted by referendum to the decision of the people. If the government

desires to enact a treaty, such as Lisbon, there needs to be a public referendum. In

2008, the first Lisbon Treaty referendum was held, and resulted in its rejection by

the electorate. The following year the Lisbon Treaty was again, put forward to the

electorate. This time it passed. There were three opportunities for the electorate to

express their preferences, and each time the result differed from the previous. The

initial vote approved of the Fianna Fáil government, the second, was a vote against

the supported position of Fianna Fáil and the third was a direct overturn of the

previous referendum decision. This paper plans to develop a greater understanding

of what happened between those elections through spatial models of the electorate

and these elections.

The analysis of these elections will center on the idea of valence for the parties

in the general election and the campaign sides in the referenda. Valence originally

proposed by Stokes (1963) is considered the non-policy attractiveness of the

candidate, or the quality of the candidate (Schofield and Gallego 2011; Gallego and

Schofield 2013). These elections will provide two different analysis of valence. One

type of analysis is the role of party valence and how that determines the ideological

position a party takes. The results from the 2007 general election illustrate how the

lowest valence party, in this case the Progressive Democrats, establish a position

that does not converge toward the electoral mean position. The second analysis

allows for a direct comparison of the valence levels of the campaign sides in the

2008 and 2009 referenda. Valence of the Yes campaign decreased between these

two referenda, and the results from the 2009 referendum show the importance of the

financial crisis in the Yes campaign winning the election.

2 2007 General Election

The 2007 election resulted in a coalition government similar to the incumbent

coalition. Fianna Fáil was still the leading party of the government, that consisted of

Fianna Fáil, the Progressive Democrats, and the Green Party. The change from the

previous coalition government was the addition of the Green party to the cabinet.

Modeling Elections and Referenda in Ireland


The election results netted Fianna Fáil and the Progressive Democrats a loss of eight

seats in the Dáil. The Progressive Democrats contributed to most of that loss, losing

six out of the eight seats. These seats were replaced with six from the Green party.

Much of the coalition program, 76 %, was taken from the Fianna Fáil manifesto,

and the Green Party was considered to have weak terms of commitment (O’Malley

2007). The Green party, a Euro-skeptic party, was particularly challenged to make

policy concessions as the government accepted a redrafted European Union treaty,

the Lisbon Treaty1 (O’Malley 2007). This treaty was then turned to the electorate in

2008 via referendum.

In order to model policy positions of the parties during the election a factor

analysis with a varimax rotation was used with 2007 data from the Irish National

Election Survey (INES, see the reference list for details on this data set)2 using the

following questions.

• It would be better if more people with strong religious beliefs held public office.

The responses were on a 5 point scale of agreement.

• Same sex marriages should be prohibited by law. The responses were on a 5 point

scale of agreement.

• People who fully agree that there should be a total ban on abortion in Ireland

would give a score of 0. People who fully agree that abortion should be freely

available in Ireland to any woman who wants to have one would give a score

of 10. Other people would place themselves in between these two views. Where

would you place yourself on this scale?

• Most of business and industry should be owned by the state : : : Most of business

and industry should be privately owned. The responses were a 10 point scale on

level of agreement.

• Where would you place yourself on these scales? The first scale is as follows

Business and industry should be strictly regulated by the state. Business and

industry should be entirely free from regulation by the state. The responses were

a 10 point scale on level of agreement.

The results from the factor analysis of the INES data, seen in Chart 1, show

that factor one includes our economic variables and factor two includes the social

variables. Using the factor loadings and individual voter’s response to the above

survey questions, we are able to estimate each voter’s position along the xi , social

position, and yi , economic position, axes.

Using these factors, the mean position (xi ,yi ) of each party’s voters are listed in

Table 1. Figure 1 presents the smoothed3 electoral distribution of the electorate’s

positions and the positions of the parties.

Using these points, we can now measure the distance of each voter, noted at

(xi ,yi ) from the positions of each party (zj ). For the purpose of this paper we will

use a pure spatial model, M(œ, “) and full model with a pure spatial model and


The Lisbon Treaty was considered a redrafted version of the European Constitutional Treaty.

INES Data made available through Irish Social Science Data Archive.


Using a Kernel Density Estimation.



N. Schofield and W. Simoneau

Chart 1 Factor analysis results for 2007 general election

Table 1 Economic and social positions of political parties





x: Social

y: Economic



















Abbreviations: FF Fianna Fáil, FG Fine Gael, GR Green, LB Labour, PD Progressive Democrats,

SF Sinn Fein, IN Independent

Fig. 1 2007 Irish electorate

sociodemographic traits M(œ, ™, “). This model assumes that the implicit utility

of voter i for party j has the below form, and that the spatial coefficient, “, must

be positive (Schofield and Gallego 2011; Schofield and Sened 2006; Gallego and

Schofield 2013).

uij xi ; zj D œij

“ k xi

zj k2 C ™j T˜i


Here ™j T ˜i models the effect of the sociodemographic characteristics ˜i of voter

i in making a political choice. Further, ™j is a vector that specifies how the various

sociodemographic variables affect choice for party (Schofield and Sened 2006).

Table 2 shows the valence estimates and spatial coefficients using a multinomial

logit model, socioeconomic coefficients can be seen in the Appendix. The results

Modeling Elections and Referenda in Ireland


Table 2 2007 General

election logit model results

œfg (intercept)

œgr (intercept)

œin (intercept)

œlb (intercept)

œpd (intercept)

œsf (intercept)




Log likelihood

LR Test

Dependent variable

Vote (spatial model)


















15.675** (df D 7)

Abbreviations: fg valence of Fine Gael,

gr valence of Green Party, in valence

of Independents, lb valence of Labour,

pd valence of Progressive Democrats,

sf valence of Sinn Fein

Note: *p < 0.1; **p < 0.05; ***p < 0.01

from the model show that the valence (œ) terms for all parties and the spatial

coefficient (“), are statistically significant in the pure spatial model. Once we

condition for demographic variables such as education, sex, and mass attendance

the spatial coefficient and all but the valence for Fine Gael (œfg ) remain significant.

Moreover, we see that voters with higher levels of education are more likely to

vote for the Greens or Labour and level of mass attendance positively affects the

likelihood of voting for Labour and Sinn Fein.4

Because of the stochastic assumption of the model, voter behavior is modeled as

a probability vector; the probability that voter i votes for party j is the probability that

the voter’s utility from voting for party j is greater than all other parties5 (Schofield

2007). Since a logit model was used we assume the cumulative distribution function,

‰, is Type I extreme value distribution6 that takes on the closed form:

‰ .h/ D exp . .exp h//


See results of full model in the Appendix.

Or, ij .z/ D Pr Œuij .xi; zj/ >. uil .xi; zl/ ; 8l Ô j.


Variance of ‰ D 16  2 .



Tài liệu bạn tìm kiếm đã sẵn sàng tải về

2 Data: The American Panel Survey (TAPS)

Tải bản đầy đủ ngay(0 tr)