Exploring Trust in Government: A Sociological Analysis Using the General Social Survey
Sebastian Joseph
2025-04-20
Introduction to Data
The point of this study is to investigate the public confidence in the federal government as measured by the General Social Survey (GSS). Having an understanding of public confidence in government is crucial because this reflect the health of our national institutions and affects civic participation, social trust, and more. From a sociological perspective, variations in confidence in government can highlight systemic inequalities, reveal divisions across demographic lines, and show the influence of economic or educational status on political attitudes. Practical and policy implications of this study include identifying groups with low trust in government and designing interventions to improve transparency, responsiveness, or outreach efforts.
Hypotheses
The dependent variable is confidence in the federal government in Washington. The independent variables are:
Age (age) - Continuous Hypothesis: Older individuals are more likely to have less confidence in the federal government. Reasoning: Older individuals are more likely to have interest in the government and its current state. This makes it more likely for older individuals to have less confidence in the federal government. Literature:
Income (coninc) - Continuous Hypothesis: Higher income is associated with more confidence in the government Rationale: Individuals with higher income may benefit more from existing government structures and policies
Political affiliation (partyid) – Categorical Hypothesis: Republicans will exhibit less confidence in the federal government when Democrats are in power, and vice versa. Rationale: Partisan alignment influences trust, with individuals trusting institutions more when their party is in control.
Education (educ) – Continuous (years of schooling) Hypothesis: Higher education levels are associated with greater confidence in government. Rationale: More educated individuals may better understand the complexities of governance and be more civically engaged.
Data and Measurement
Dependent Variable: - confed: “How much confidence do you have in the people running the federal government in Washington?” - 1 = A great deal - 2 = Only some - 3 = Hardly any
Independent Variables: - age: Respondent’s age (in years) - coninc: Constant comparative income (adjusted household income) - partyid: “Generally speaking, do you usually think of yourself as a Republican, Democrat, Independent, or what?” - educ: Respondent’s years of schooling completed
Here’s the construction of the gss and gss confidence data frame
library(tidyverse)## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.4 ✔ readr 2.1.5
## ✔ forcats 1.0.0 ✔ stringr 1.5.1
## ✔ ggplot2 3.5.2 ✔ tibble 3.2.1
## ✔ lubridate 1.9.4 ✔ tidyr 1.3.1
## ✔ purrr 1.0.4
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorsgss_survey <- read.csv("C:/Users/Thecl/Desktop/Kean/SPR2025/MATH3710/FInal_Project/gss_3325.csv")
data(gss_survey)## Warning in data(gss_survey): data set 'gss_survey' not foundgss_conf = gss_survey %>%
select(year, confed, age, coninc, partyid, educ) %>%
filter(!is.na(confed) & !is.na(age) & !is.na(coninc) & !is.na(partyid) & !is.na(educ))Univeriate Analysis
# Frequency of confidence in federal government
table(gss_conf$confed)##
## A Great Deal Hardly Any Only Some
## 6707 14579 21405Confidence in the US Government
The variable confed (Confidence in the US Government) measures how much confidence individuals have in the federal government. The distribution is as follows:
Age
# Summary of age
summary(gss_conf$age)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 18.0 31.0 43.0 45.8 59.0 89.0ggplot(gss_conf, aes(x = age)) +
geom_histogram(binwidth = 5, fill = "darkorange", color = "white") +
labs(title = "Age Distribution of Respondents",
x = "Age",
y = "Count") +
theme_minimal()
This data shows that only about 1 in 6 americans have reported high confidence in the federal government, with the majority expressing moderate or low trust in the government. This data highlights a potentially concerning level of public skeptism.
Income (CONINC)
# Summary of income
summary(gss_conf$coninc)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 336 18480 36482 45748 59895 180386ggplot(gss_conf, aes(x = coninc)) +
geom_histogram(binwidth = 5000, fill = "forestgreen", color = "white") +
labs(title = "Household Income Distribution",
x = "Income (USD)",
y = "Count") +
theme_minimal()
These respondents span a wide age range which is providing a good basis for testing how age influences political trust.
Education
gss_conf$educ_clean <- gss_conf$educ
gss_conf$educ_clean[gss_conf$educ_clean == "No Formal Schooling"] <- "0"
gss_conf$educ_years <- as.numeric(gss_conf$educ_clean)
# Summary of education
summary(gss_conf$educ_years)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.00 12.00 12.00 13.04 16.00 20.00table(is.na(gss_conf$educ_years))##
## FALSE
## 42691gss_conf$educ_clean <- gss_conf$educ
gss_conf$educ_clean[gss_conf$educ_clean == "No Formal Schooling"] <- "0"
gss_conf$educ_years <- as.numeric(gss_conf$educ_clean)
ggplot(gss_conf, aes(x = educ_years)) +
geom_histogram(binwidth = 1, fill = "purple", color = "white") +
labs(title = "Years of Education Completed",
x = "Years of Schooling",
y = "Count") +
theme_minimal()
Income shows substantial variability, indicating that respondents come
from a wide range of economic backgrounds, which is helpful for
evaluating the relationship between socioeconomic status and trust in
government.
Political Affiliation
Party identification is quite diverse among respondents:
# Frequency of political identification
table(gss_conf$partyid)##
## Independent (neither, No Response) Independent, Close To Democrat
## 6106 5465
## Independent, Close To Republican Not Very Strong Democrat
## 3998 8876
## Not Very Strong Republican Other Party
## 6447 731
## Strong Democrat Strong Republican
## 6982 4086ggplot(gss_conf, aes(x = fct_infreq(partyid))) +
geom_bar(fill = "dodgerblue4") +
coord_flip() +
labs(title = "Party Identification",
x = "Party ID",
y = "Count") +
theme_minimal()
Bivariate Analysis
Age vs Confidence in Government
library(dplyr)
gss_conf = gss_conf %>%
mutate(age_group = cut(age, breaks = c(17, 30, 45, 60, 90),
labels = c("18-30", "31-45", "46-60", "61+")))
age_conf_table <- table(gss_conf$age_group, gss_conf$confed)
prop.table(age_conf_table, margin = 1) * 100##
## A Great Deal Hardly Any Only Some
## 18-30 15.95918 30.57143 53.46939
## 31-45 14.37230 33.92681 51.70089
## 46-60 15.51553 35.74841 48.73607
## 61+ 17.49274 36.42679 46.08046The percentages of respondents with “Hardly Any” confidence increased with age, peaking in the 61+ group. The youngest group (18-30) is more likely to respond with “Only Some”. Trust has shown to decrease slightly with age but older respondents are also more polarized.
Income vs Confidence in Government
gss_conf <- gss_conf %>%
mutate(income_quartile = ntile(coninc, 4))
income_conf_table <- table(gss_conf$income_quartile, gss_conf$confed)
prop.table(income_conf_table, margin = 1) * 100##
## A Great Deal Hardly Any Only Some
## 1 15.89993 35.58512 48.51494
## 2 15.37525 33.50511 51.11965
## 3 15.21597 34.54511 50.23892
## 4 16.35120 32.96477 50.68403Higher income groups are slightly more likely to say “A Great Deal” and less likely to say “Hardly Any”. Overall, confidence is relatively stable across income, but the wealthiest show a small upward trend in trust.
Education vs Confidence in Government
educ_conf_table <- table(gss_conf$educ_years, gss_conf$confed)
prop.table(educ_conf_table, margin = 1) * 100##
## A Great Deal Hardly Any Only Some
## 0 23.52941 33.82353 42.64706
## 1 11.11111 25.92593 62.96296
## 2 10.66667 36.00000 53.33333
## 3 28.00000 26.66667 45.33333
## 4 25.14286 30.28571 44.57143
## 5 23.42342 35.13514 41.44144
## 6 19.91952 28.57143 51.50905
## 7 20.36697 30.82569 48.80734
## 8 18.67728 29.33252 51.99020
## 9 15.81749 35.51331 48.66920
## 10 16.53072 35.23654 48.23274
## 11 15.31842 34.38038 50.30120
## 12 14.14306 34.45580 51.40115
## 13 14.87819 35.64385 49.47796
## 14 14.27356 37.41757 48.30887
## 15 15.91963 36.26996 47.81041
## 16 16.84810 31.88746 51.26444
## 17 18.38596 31.36842 50.24561
## 18 15.83236 35.38999 48.77765
## 19 17.68202 30.01486 52.30312
## 20 17.54386 33.52827 48.92788This suggests a U-shaped trend, where both low and high education levels show greater trust, while middle education (e.g., some high school or only high school) is more skeptical.
Party ID vs Confidence in Government
party_conf_table <- table(gss_conf$partyid, gss_conf$confed)
prop.table(party_conf_table, margin = 1) * 100##
## A Great Deal Hardly Any Only Some
## Independent (neither, No Response) 11.12021 39.97707 48.90272
## Independent, Close To Democrat 12.13175 35.64501 52.22324
## Independent, Close To Republican 15.13257 36.74337 48.12406
## Not Very Strong Democrat 14.53357 30.65570 54.81073
## Not Very Strong Republican 17.35691 30.20009 52.44300
## Other Party 10.67031 51.16279 38.16689
## Strong Democrat 18.53337 32.49785 48.96878
## Strong Republican 23.95986 34.50808 41.53206Strong Republicans are most likely to say “A Great Deal.” Independents and other parties show the least confidence—with 40–50% saying “Hardly Any.” Partisan identity is highly predictive of confidence—strong partisans tend to trust more, independents trust less.
Multinomial Logistic Regression
library(nnet)
model_data <- gss_conf %>%
select(confed, age, coninc, educ_years, partyid) %>%
na.omit()
model_data$confed <- relevel(as.factor(model_data$confed), ref = "Only Some")
model_multinom <- multinom(confed ~ age + coninc + educ_years + partyid, data = model_data)## # weights: 36 (22 variable)
## initial value 46900.857216
## iter 10 value 43035.435740
## iter 20 value 42820.065527
## final value 42439.053859
## convergedsummary(model_multinom)## Call:
## multinom(formula = confed ~ age + coninc + educ_years + partyid,
## data = model_data)
##
## Coefficients:
## (Intercept) age coninc educ_years
## A Great Deal -1.5087862 0.002899610 1.213047e-08 -0.007437429
## Hardly Any -0.5940222 0.007293365 -7.678845e-07 0.009183075
## partyidIndependent, Close To Democrat
## A Great Deal 0.02477172
## Hardly Any -0.18974728
## partyidIndependent, Close To Republican
## A Great Deal 0.32284692
## Hardly Any -0.08631721
## partyidNot Very Strong Democrat partyidNot Very Strong Republican
## A Great Deal 0.1470436 0.3700041
## Hardly Any -0.3984453 -0.3751190
## partyidOther Party partyidStrong Democrat partyidStrong Republican
## A Great Deal 0.2133667 0.4852954 0.9099513
## Hardly Any 0.4857859 -0.2726594 -0.0508477
##
## Std. Errors:
## (Intercept) age coninc educ_years
## A Great Deal 6.427945e-06 0.0003983676 3.426631e-07 7.156937e-05
## Hardly Any 4.983301e-06 0.0003060581 2.675115e-07 5.618077e-05
## partyidIndependent, Close To Democrat
## A Great Deal 5.978557e-07
## Hardly Any 6.372895e-07
## partyidIndependent, Close To Republican
## A Great Deal 4.603514e-07
## Hardly Any 3.889360e-07
## partyidNot Very Strong Democrat partyidNot Very Strong Republican
## A Great Deal 1.298601e-06 8.273400e-07
## Hardly Any 1.047035e-06 5.496481e-07
## partyidOther Party partyidStrong Democrat partyidStrong Republican
## A Great Deal 5.610586e-08 1.699024e-06 8.152051e-07
## Hardly Any 7.383508e-08 1.080606e-06 4.217016e-07
##
## Residual Deviance: 84878.11
## AIC: 84922.11z <- summary(model_multinom)$coefficients / summary(model_multinom)$standard.errors
p_values <- 2 * (1 - pnorm(abs(z)))
p_values## (Intercept) age coninc educ_years
## A Great Deal 0 3.370637e-13 0.971760334 0
## Hardly Any 0 0.000000e+00 0.004098583 0
## partyidIndependent, Close To Democrat
## A Great Deal 0
## Hardly Any 0
## partyidIndependent, Close To Republican
## A Great Deal 0
## Hardly Any 0
## partyidNot Very Strong Democrat partyidNot Very Strong Republican
## A Great Deal 0 0
## Hardly Any 0 0
## partyidOther Party partyidStrong Democrat partyidStrong Republican
## A Great Deal 0 0 0
## Hardly Any 0 0 0All p-values are 0 or very close to 0, which means the following variables are statistically significant predictors of confidence level:
Age: Positive for both “A Great Deal” and “Hardly Any”. This suggests that as age increases, people are more likely to express either extreme of confidence compared to “Only Some”
Income: Positive for “A Great Deal”, Negative for “Hardly Any”. It suggests that igher-income individuals are more likely to have greater trust in the federal government.
Education: Negative for “A Great Deal”, Positive for “Hardly Any”. Suggests more years of schooling correlates with less confidence in government. Could reflect critical thinking or systemic disillusionment in higher-educated respondents.
Party ID: Strong republicans and Strong Democrats are more likely to express “A Great Deal of Confidence”. Not Very Strong Democrats, Other Party, and Close to Democrat are more likely to show “Hardly Any” confidence. Strong Democrats are less likely to be in “Hardly Any” category.
These effects reflect deep political divides and partisanship’srole in shaping instituitional trust.
Diagnostics Summary
- Residual Deviance: 84878.11 AIC: 84922.11
Conclusion
This study investigated the factors influencing public confidence in the U.S. federal government using information from the General Social Survey ( GSS ). The results reveal that trust in the government is not uniformly distributed across the population but is significantly shaped by age, income, education, and political affiliation.
Age emerged as a significant predictor, alongside older individuals who were more likely to express either strong confidence or strong distrust, suggesting polarization increases alongside age. Income had a modest positive relationship with confidence: higher-income individuals were slightly more likely to trust the government, consistent together with the hypothesis that economic benefit may influence perceptions of institutional effectiveness.
Interestingly, education showed a counterintuitive pattern: higher education levels were associated together with less trust in government. This finding contradicts the original hypothesis that may reflect a more critical or skeptical perspective among the highly educated, perhaps due to increased awareness of systemic shortcomings.
Political identity was the strongest along with the most consistent predictor. Strong partisans — particularly Strong Republicans as well as Strong Democrats — were significantly more likely to express high levels of confidence. In contrast, Independents as well as minor party affiliates were more likely to report low confidence, suggesting which partisan alignment plays a key role in shaping political trust.
Overall, the findings support the idea that political trust is not only a function of personal resources like income and education but is deeply rooted within political identity in addition to broader sociopolitical narratives. These insights have implications for efforts to rebuild public trust, suggesting which strategies must address not only transparency as well as performance but also the political climate as well as how citizens perceive their relationship with institutions.
Citations
Citrin, J., & Stoker, L. (2018, May 11). Political Trust in a cynical age. Annual Review of Political Science. https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-050316-092550
Weinschenk, A. C., & Helpap, D. J. (2015). Political Trust in the American States. State and Local Government Review, 47(1), 26-34. https://doi.org/10.1177/0160323X15569547
Hunt, N.C., Iyer, G.S. and Jimenez, P. (2019), Election Outcome and Tax Compliance: The Role of Political Party Affiliation, Affect Balance, and Trust in Government. Applied Psychology, 68: 341-372. https://doi.org/10.1111/apps.12165
Ugur-Cinar, M., Cinar, K. & Kose, T. How Does Education Affect Political Trust?: An Analysis of Moderating Factors. Soc Indic Res 152, 779–808 (2020). https://doi.org/10.1007/s11205-020-02463-z