POLS GU4712 HW 4

1. Load + Organize Data

# Load data 
pmeans <- read.csv("/Users/linde/Downloads/LavBen_Germany.csv")

# Keep just the numeric columns subset so I can run fa.parallel
pmeans_num <- pmeans[, c("Mean1", "Mean2", "Mean5", "Mean12",
                         "Mean17", "Mean19", "Mean23", "Mean24")]

pmeans_num <- as.data.frame(sapply(pmeans_num, as.numeric))

2. Descriptions of Select Mean Positions

Issue Area 1 (spending vs. taxes): A 20-point scale measuring whether the mean party position leans toward raising taxes to increase public services (1) or cutting public services to cut taxes (20).
Issue Area 2 (liberal policies): A 20-point scale measuring whether the mean party position favors liberal policies towards abortion, homosexuality, and euthanasia (1) or opposes those policies (20).
Issue Area 5 (environment): A 20-point scale measuring whether the mean party position privileges environmental protections over economic growth (1) or privileges economic growth over environmental protections (20).
Issue Area 12 (decentralization): A 20-point scale measuring whether the mean party position favors policies decentralizing all administrative responsibilities (1) or opposes those policies (20).
Issue Area 17 (EU peacekeeping): A 20-point scale measuring whether the mean party position favors German involvement in European security and peacekeeping missions (1) or opposes such involvement (20).
Issue Area 19 (immigration to the EU): A 20-point scale measuring whether the mean party position leans toward promoting safe immigration to Germany (1) or promoting the return of immigrants to their countries of origin (20).
Issue Area 23 (EU accountability): A 20-point scale measuring whether the mean party position leans toward promoting the direct accountability of the EU to citizens via multinational institutions (1) or promoting the indirect accountability of the EU to its citizens via national governments (20).
Issue Area 24 (EU authority): A 20-point scale measuring whether the mean party position leans towards increasing the EU’s policy jurisdiction (1) or decreasing the EU’s policy jurisdiction (20).

3. Descriptions of Political Parties + Popular Support Pre-2002

Information on the political parties taken from the respective Encyclopedia Britannica pages for each party.* Information on popular support during and prior to the 2002 elections taken from: this CRS report; this journal article by Peter James; and Chapter 1 of this book by David P. Conradt, which is not available through Columbia Libraries but which I accessed through my Sciences Po account.

* The exceptions are the DKP, DVU, and PDS, for which there are no Encyclopedia Britannica pages. Instead, I used this page from Wikiwand for the DKP (questionable website name, but its info is good), this LOC study for the DVU, and the Wikipedia page (I know, I know) for the PDS.

Christian Democratic Union/Christian Social Union (CDU-CSU): The CDU and its Bavarian affiliate with whom it always coalitions, the CSU, is a center-right party that supports a free market economy, federalism, and social welfare programs while maintaining conservative stances on social issues. The CDU-CSU suffered a major loss to the SPD in the 1998 election, which scholar Peter James attributes to its inability to adequately address high unemployment and reunification stagnation.
German Communist Party (DKP): The DKP is a fringe party that has enjoyed small success chiefly in working class and university towns. It upholds traditional communist values of redistribution and workers’ rights. Its popularity decreased rapidly following reunification, and it has never held any seats in the Bundestag.
German People’s Union (DVU): The DVU was a far-right, extremely nativist party that espoused many of the views held by the Republicans but went a step further by tacitly supporting violence against immigrants. In April 1992, Germany’s domestic intelligence agency announced that the DVU was under surveillance to determine if it needed to be banned for being anti-democratic.
Free Democratic Party (FDP): The FDP is a centrist party that supports a free market economy and social reform. From its inception in 1948 to our focus year, 2002, the FDP shifted from coalition to coalition, either with the CDU-CSU or the SPD. In the early 1990s, the FDP failed to win representation in several state elections, and so after reassessing its policies chose to emphasize environmental protection (to combat the Green Party), assertive foreign policy, and conservative economic policy.
Green Party (GRÜ): The GRÜ emphasizes environmental protection, specifically from nuclear technologies, and nonviolence. While the GRÜ rose in popularity after the 1986 Chernobyl nuclear disaster, it did not win representation in the 1990 elections, and so allied with the East German Greens and a coalition of grassroots organizations known as Alliance ’90. The Greens formed coalition with the SPD in the 1998 election, which they won, but which was bittersweet since some of the actions taken by the government, including sending forces to Kosovo and Afghanistan, went against their principle of nonviolence.
National Democratic Party (NPD): The NPD, now known as “The Homeland”, is a far-right, ultranationalist, isolationist, and economically conservative neo-Nazi party. It achieved moderate electoral success in the 1960s due to fears over immigration, but has never held any seats in the Bundestag. Its popularity leading up to the 2002 election was low.
Party of Democratic Socialism (PDS): The PDS (predecessor to the contemporary Die Linke) was a left-wing populist party that was very popular in the eastern states due to it being the successor of the leading governing party of East Germany. The PDS officially rejected any Marxist-Leninist policies and promoted socially progressive ones, such as legalizing same-sex marriage. It did relatively well in the 1998 election, winning seats in the Bundestag.
Republicans (Rep): The Rep is a far-right neo-fascist party that is staunchly opposed to immigration. Like the DVU, the Rep was being monitored by the German domestic intelligence agency at the time of the 2002 elections. Throughout the 1990s and early 2000s, there were divisions within the Rep about the role of extremism in the party’s ideology and alliances, and support for the party steadily declined.
Social Democratic Party of Germany (SPD): The SPD is the main competitor of the CDU-CSU. It supports economic modernization and social welfare. In 1998, the SPD-Green coalition unexpectedly won a strong majority in the Bundestag, signaling a new, more left-leaning era of German politics.
Partei Rechtsstaatlicher Offensive (Schill): Founded in 2000, the Schill party took its name from its founder, Ronald Schill, who was a former judge and saw the current political system as insufficient for addressing crime. Because of this, its ideologies emphasize harsher punishments for crimes and are generally more conservative. It received support in the 2001 Hamburg elections due to anxieties over 9/11.

4. Scree Plot + Eigenvalues

parallel <- fa.parallel(pmeans_num, fm = 'minres', fa = 'fa')

## Parallel analysis suggests that the number of factors =  2  and the number of components =  NA

In conclusion:

This model (and also the models with 3 and 4 factors) are terrible. TLI isn’t even close to 0.9, and the RMSEA is really high. But, per what you said in office hours, this is because there are only 10 observations. So I continue on.
The eigenvalues for the “first few” (I am including 4) factors are as follows: 3.18; 1.84; 1.61; and 1.15. But, this changes depending on how many factors you include in the model. In office hours, you told me to eyeball the eigenvalues based on the scree plot (as opposed to just including the values obtained from factor analysis, which I listed above), so those values for the first two factors would be 5.00 and 1.50, respectively.
According to this data, German politics has 2 dimensions. I reached this conclusion by looking at the scree plot, which shows only 2 factors with eigenvalues greater than 1 and above the simulated data line. Additionally, the scree plot output explicitly says that the suggested number of factors is 2.

5. Factor Analysis Model (3 factors)

threefactor <- fa(pmeans_num, nfactors=3, rotate="oblimin", fm="minres")
print(threefactor)

## Factor Analysis using method =  minres
## Call: fa(r = pmeans_num, nfactors = 3, rotate = "oblimin", fm = "minres")
## Standardized loadings (pattern matrix) based upon correlation matrix
##          MR1   MR2   MR3   h2      u2 com
## Mean1   0.29 -0.93  0.21 0.98  0.0178 1.3
## Mean2   1.00 -0.02 -0.03 0.95  0.0514 1.0
## Mean5  -0.04 -0.01  1.02 1.00  0.0043 1.0
## Mean12  0.16  0.89  0.16 0.91  0.0855 1.1
## Mean17  0.45  0.69  0.03 0.84  0.1581 1.7
## Mean19  0.99 -0.05  0.03 1.00 -0.0015 1.0
## Mean23  0.92 -0.07  0.10 0.97  0.0300 1.0
## Mean24  0.70  0.41  0.13 0.94  0.0577 1.7
## 
##                        MR1  MR2  MR3
## SS loadings           3.89 2.38 1.32
## Proportion Var        0.49 0.30 0.17
## Cumulative Var        0.49 0.78 0.95
## Proportion Explained  0.51 0.31 0.17
## Cumulative Proportion 0.51 0.83 1.00
## 
##  With factor correlations of 
##      MR1   MR2   MR3
## MR1 1.00  0.22  0.73
## MR2 0.22  1.00 -0.07
## MR3 0.73 -0.07  1.00
## 
## Mean item complexity =  1.2
## Test of the hypothesis that 3 factors are sufficient.
## 
## df null model =  28  with the objective function =  21.77 with Chi Square =  119.74
## df of  the model are 7  and the objective function was  6.84 
## 
## The root mean square of the residuals (RMSR) is  0.02 
## The df corrected root mean square of the residuals is  0.05 
## 
## The harmonic n.obs is  10 with the empirical chi square  0.15  with prob <  1 
## The total n.obs was  10  with Likelihood Chi Square =  23.93  with prob <  0.0012 
## 
## Tucker Lewis Index of factoring reliability =  -0.405
## RMSEA index =  0.48  and the 90 % confidence intervals are  0.3 0.753
## BIC =  7.81
## Fit based upon off diagonal values = 1
## Measures of factor score adequacy             
##                                                    MR1  MR2  MR3
## Correlation of (regression) scores with factors   0.99 0.99 0.99
## Multiple R square of scores with factors          0.98 0.98 0.99
## Minimum correlation of possible factor scores     0.96 0.96 0.97

6. Factor Analysis Interpretations

Factor 1 loads strongly on Mean2 (liberal social policies), Mean19 (immigration), Mean23 (EU accountability), and Mean24 (EU authority). Lower scores on these survey questions correspond to more liberal, pro‑immigration, and pro‑integration positions, whereas higher scores correspond to conservative, anti‑immigration, and Euroskeptic positions. Because parties with traditionally left, cosmopolitan, pro‑EU platforms (e.g., GRÜ, PDS, SPD) tend to have low values on these items, while far‑right or nationalist parties (e.g., DVU, NPD, Rep, Schill) tend to have high values, I interpret low Factor 1 scores as “socially-left/cosmopolitan” and high Factor 1 scores as “socially-right/nationalist”.
Factor 2 loads strongly on Mean1 (economic policy), Mean12 (decentralization), and Mean17 (EU peacekeeping). Lower scores on these survey questions correspond to more economically-liberal (with respect to raising taxes for public welfare, sharing responsibilities for financial decision-making, and allocating funds towards international peacekeeping efforts) whereas higher scores correspond to more economically-conservative and isolationist positions. Because parties with traditionally left, economically-liberal platforms (e.g., GRÜ, CDU-CSU, FDP) tend to have low values on these items, while far‑right, isolationist, or economically-alternative (e.g., communist) parties (e.g., DVU, NPD, DKP) tend to have high values, I interpret low Factor 2 scores as “economically-left/market reformists and high Factor 2 scores as”economically-right/market alternativists.

7. Plotting

plot(threefactor$scores[,1], threefactor$scores[,2], pch=16, col="red", xlab="Factor 1", ylab="Factor 2", main="Factor Analysis of German Political Parties in 2002", xlim=c(-2.5, 2.5), ylim=c(-2.5, 2.5)) 
text(threefactor$scores[,1], threefactor$scores[,2], labels=c("CDU-CSU", "DKP", "DVU", "FDP", "GRÜ", "NPD", "PDS", "Rep", "SPD", "Schill"),, pos = 4)
mtext("Plot by Linden James using data from Laver and Benoit", side = 3, line = 0.5, cex = 0.8)

8. Interpretations + Comparisons

The two main parties, SPD and CDU-CSU, are relatively close together. This makes sense, as they share very similar ideologies, such as promoting economic modernization through the free market economy and supporting social welfare initiatives. Perhaps one takeaway could be that if one party is very prominent and another party shares much of that party’s ideology, then that other party would also be prominent. However, Rep is also relatively close to both CDU-CSU and SPD, without sharing many of their ideologies and without being very prominent. Perhaps one reason for this clustering, then, is that, within each left-right placement, parties that hover more towards the center (0 for Factor 1 and 0 for Factor 2) are more likely to be prominent. This makes general sense, as political and economic centrism seemed to be a prevailing ideology characteristic of Germany in the late 1990s-early 2000s.
Party endurance/prominence can be visualized: generally, parties scoring lower on Factor 2 (specifically, under 0) have undergone fewer transitions/have not been banned/disbanded and have a solid amount of seats in the Bundestag. This builds on the conclusion forwarded in the above paragraph to suggest that parties that are more prominent, mediated by their centrist tendencies, are more likely to endure in their current form. This makes sense: it should be harder for parties with lots of power and influence to fall out of favor than it should be for parties with limited power and influence.