### Introduction

Since the mid 60s, political trust within the United States has plummeted from 77% to just 17% as of 2019. Within that same time span, we have witnessed the end of the Democratic hegemony, the establishment of the modern Republican party, and the rise to supremacy of the Independent affiliation.While the two phenomena may appear to be independent of each other, I believe them to be greatly intertwined. In this article, I will be discussing the role that political trust plays in partisanship identity, both in terms of formation and strength.I will then conclude by presenting early evidence that this same paradigm might play a role in partisan sorting.Though the study continues, and no causal conclusions have been made, the findings of this article are very promising.

### The Psychological Contract

The theory proposed by this article revolves around the concept of a psychological contract, first proposed by Dr. Argyris in 1960. To put into laymans terms, there exists a psychological contract between an employer and employee, defined by the mutual expectations present upon hire. For example, the employer might expect that it is agreed that the employee is better than the last hire, does his job efficiently, boosts morale, and presents well to the clientele. Likewise, the employee might expect that the employer pays them “fairly” for their work, is not “oppressive”, offers “acceptable” working conditions, and respects workers rights. These expectations are considered mutually accepted upon hire, and form an unwritten contract (psychological) between the organization/employer and the employee.

However, when that contract is violated by either party, trust is lost. This is critical with regards to organizational citizenship behavior (OCB), and other factors, such as loyalty and commitment.When trust in the organization is depleted, so with it goes commitment, loyalty, and organizational citizenship behavior (OCB). Now, you might be wondering what OCB is. Organizational citizenship behavior is simply discretionary behavior which is not part of the specified individuals job description but is believed to benefit the organization. This could include a sociable attitude, which promotes morale, or helping to guide a new hire. You can hopefully understand how the violation of these psychological contracts could be detrimental to businesses and their employees. On the employer’s end, you could observe terminations, demotions, a cut in hours, or even general hostility (for both sides). On the employees end, there could be less efficiency, lower morale, higher turnover, or lower client satisfaction.

### An Interdisciplinary Theory

##### Political Science + Organizational Behavior

Our political system very much resembles a corporate organization, both in terms of structure and behavior. We can view the political party as the aggregate organization, and their representatives as the administration. Furthermore, we may view the voter as an employee, as they are indoctrinated members of the organization, and vote for the candidates, which can be conceptualized as labor because political participation takes effort.Financial incentive, in this case, can be conceptualized as legislative outcomes.

A voter has obtained expectations of their affiliated party, which was probably set in place through socialization and conditioning by their sphere of influence. Nevertheless, they have formed a psychological contract with the party. The voter is expected to participate in elections, such as turning out, following political news, donating, or even IDENTIFYING with the party. The organization, or political party, is expected to fulfill campaign promises, maintain party ideology, and promote partisan agendas.

If the party violates the psychological contract, thus betraying the voter’s expectations, then the consequence is a loss of trust. When this occurs, the voter is expected to protest, or retaliate for lack of a better word. They might not turnout in the upcoming election, donate to a campaign, or place a bumper sticker on their car. However, if the betrayal is severe enough, it is the theory of this article that the voter might react by leaving the organization altogether. This, of course, is dependent upon the individual.

However, this does not mean that the voter will jump ship. No, they are still partisans, with the same preferences that originally incentivized their membership. Instead, due to the loss of trust and accompanying frustration/anger, the voter will proclaim independence. It is therefore no coincidence why we have witnessed both a dramatic decrease in political trust, alongside a stunning rise in independent affiliation.

### Aggregate Analysis

##### Trust and Independence

The initial aggregate level regression resulted in very strong findings. By controlling for events, such as Vietnam, economic contractions/expansions, war bumps, and presidential elections, I was able to show the power of trust on independent affiliation. The reason why these events were controlled for has to do with marginal affect, which will be discussed shortly. These are periods in time that produced bi-partisan reaction, and added to the independent affiliation by numbing the effect of partisanship.Even the 80s saw a bipartisan rise in political trust, although most of the addition came from the Republicans during the Reagan presidency.

### Individual Level Analysis

##### Trust & Independence

At the individual level, given my theory, one would expect that as a voter moves from the highest level of trust to the lowest, their likelihood of identifying as an independent would increase. This is exactly what the second model shows us. However, I went a step further, and showed that trust not only dictates general partisan identity, it also effects its strength. Going from 1-7 on the partisan intensity scale (Strong Dem – Strong Rep), if my theory holds true, then you should see a horseshoe pattern in the margins plot. Observing the first model, it would seem as though that is the case.

### Marginal Affect:

##### The Difference Maker

When I began to look at the individual level data, the initial results were significant, but sloppy. There was something occurring at the individual level that the aggregate was able to control for with the mean. Upon further contemplation, I decided to look at how affection towards your own party could affect the models outcome. Sure enough, it boosted the results, and cleaned up the margins plot quite a bit. However, there was still work to be done. I carefully considered what was taking place when moving from aggregate to individual, realizing that it is not simply just how much you might like or dislike your own party. Rather, it is the difference between the affect for ones own party, and the affect (love/hate/indifference) for the opposite. The reason is that an individual who truly loves his team and hates the other team would probably react differently to a drop in trust, compared to an individual who likes his team and is indifferent to the other. This strength of affection is critical in how a voter might react in a situation where the psychological contract was violated. So, I simply computed the margins between the two, and created the variable “marginal affect”. When added to the logistic regression, everything tightened up (statistically speaking). The results are pictured in the following:

### Voter Registration Confirmation

##### Surprising Results

I wanted to see if trust had the same effect on voter registration as it did self-reported affiliation. So, I gathered data from the North Carolina voter file, looking at registration dates, and ran a simple regression. The results are exactly as I had hoped.As you can see, by simply comparing these results to the initial analysis previously discussed, political trust has a major influence over independent registration as well as self reported affiliation. The two are almost identical.

Next, I wanted to compare correlation results. How does each party compare to one another in terms of self-reported versus registration percentages? I wanted to conduct this analysis because it would let me know if the three classes of partisans within North Carolina behave similarly (or dissimilarly) to the overall population. This would let me know if the regression results were valid. As you can see, they were. For the most part that is.

As you can see, the Independents (.9065) and Democrats (.9072) were highly correlated. However, the Republicans (.55) did not fare as well.Though, this is exactly what I was hoping to find. Though a bit complicated, I will attempt to explain my logic as clearly as possible.

I believe that the DEMs in North Carolina have been greatly weakened since the 60`s. Being that the slope of the Independents is suspiciously similar, and the two are both share a strong negative correlated, I would have to suspect that the Democrats in North Carolina have been let down by their party/organization. They had their psychological contract violated, which resulted in a loss of political trust. This led many to protest, seeking independent affiliation.

If you look at the correlation table, you will notice that the Republican registration percentage correlates just as well with the Democratic self-reported affiliation as it does with its own self-reported affiliation. Further inspection led to the observation that it is also correlated with Democratic registration (-.45) and Independent registration (-.89) in North Carolina. What this leads me to believe is that the Republicans have been losing out to the Democrats indirectly, as they have been racing towards independent registration within North Carolina.

What I believe this shows is that my theory of independent affiliation helps explain partisan sorting. The independent affiliation is a vesicle for disillusioned partisans, and the variation of party percentages over the years is a result of these members leaving their party for independent affiliation when their psychological contracts have been violated, and returning once trust is replenished. Much more analysis needs to be done to validate my claims, but I am hopeful.

It also leads me to believe that location is critical when studying the affects of political trust on partisan ID. I will certainly be adding the state of an individual, and might even go as far as to add their residing county. Maybe partisan ID is viral, like a social trend. Location, if densely leaning, could inspire partisan ID, especially concerning Independent affiliation, which is more of a statement than an identity.Yes, much more needs to be done here, including a better measurement of political trust.

### Data Solutions

Really nothing too complex here. There were a few bumps in the road, but I managed to divert them. The first was the dates, which came in the form of yyyy/mm/dd. Since I was interested in the month and day, I had to separate them. I used the following code:

gen Month_of_Register=substr(registr_dt,6,2)

gen Year_of_Register=substr(registr_dt,1,4)

gen Day_of_Register=substr(registr_dt,9,2)

• By completing this, I was able to [tab Party_of_Register Year_of_Register if Year_of_Register==X] for example.

Next was the variables I had to create. The first was marginal affect, which is the difference between two differences (like-dislike)-(like-dislike). Next, I had to generate dummy variables for various events. This was achieved with a simple gen X / if y. Lastly, I had to create the registration percentage which, again, was simple arithmetic. Below is the formula used for “marginal affect”.

1. Marginal Affect • sqrt(DemAffect-RepAffect)^2 Along the way I did have to clean up some data. Since North Carolina, PEW, and ANES are all very thorough, this was not a very difficult task.

• numeric from string [destring party_to, generate(Party_To)]

• labeled numeric from string [encode county_name, generate(County)]

• ordered numeric (ordinal) instead of randomly assigned [label define Partisan_Switch 1 “DEM/UNA” 2“DEM/REP” 3 “LIB/UNA” 4 “REP/UNA” 5 “UNA/REP” 6 “LIB/DEM” 7 “REP/ DEM” 8 “UNA/DEM” 9 “UNA/CST”, replace]

• Creation of dummy variable using (replace X= , if Y== ) , and also had to use the same function to get rid of 0 where a ( – ) should be, using (replace X= . , if Y== .)

Importing the data posed some difficulties due to the size of the data sets, but I managed to get around this. Images also had to be transformed into .png, including tables, before I could import to R markdown. Lastly, I had to shrink the ANES data sets, so they were more manageable.

• Had to limit the data set to only relevant columns since the file was so large. By doing so, I was able to download every observation at once instead of in segments. It also transforms the .txt file. Example: __import delimited "C:G Latimer_20150101.txt", delimiter(tab) varnames(1) colrange(44:47)__

• The images used in the article were transformed into .png files using an image editor. They were initially STATA produced images.

• I had to shrink the ANES data sets by deleting variables, so it was more manageable.

Example: [drop age birth_place registr_dt]

• Importing Images to R- EXC{}(/Users/John G Latimer/Desktop/EEE.png)))

Lastly was the creation of the tables and graphs. This includes the regression tables, correlation table, margins plots, and two-way graphs. There were really no difficulties here, as the commands are fairly straightforward.

1. Margins Plots
1. logistic Independent Trust Marginal_Affect Age i.Gender i.Race Education i.Religion i.Mother_Party i.Father_Party Registered

2. margins, at (Trust=( 1 2 3 4))

3. marginsplot

1. Two Way Graphs
1. twoway (line dem_reg_perc year) (line demaffiliation year)
1. Table of Graphs
1. graph combine “C:G Latimerand Trust PRedict.gph” “C:G LatimerAffect and IND.gph”