1 Introduction

This script analyzes how social determinants—namely income satisfaction, confidence in health services, childhood conflicts, and working hours—are associated with depressive symptoms in Austria, using data from the European Social Survey (ESS Round 11, 2022).

Depression is measured with the CES-D8 scale, which captures the frequency of depressive experiences in the past week. The analysis includes both continuous and binary versions of the CES-D8 index and is structured as follows:

Recoding of CES-D8 items
Reliability analysis of the scale
Descriptive statistics and visual exploration
Bivariate and multivariate modeling
Interpretation of predictors of clinically relevant depression
Presentation of key results in visual form

2 Recoding the CES-D8 Depression Scale

The CES-D8 scale is composed of eight items that reflect core depressive symptoms. Responses were originally categorical and have been recoded to numeric values from 0 to 3, where higher values indicate more frequent symptoms. Two positively phrased items (happiness and enjoyment of life) were reverse-coded to ensure consistency. A total score (CESD_TOTAL) was calculated by summing the recoded items. This continuous variable serves as the main dependent variable throughout the analysis.

cesd_items = c("fltdpr", "flteeff", "slprl", "wrhpp", "fltlnl", "enjlf", "fltsd", "cldgng")
for (item in cesd_items) {
  df_aus[[paste0(item, "_n")]] <- NA
  df_aus[[paste0(item, "_n")]][df_aus[[item]] == "None or almost none of the time"] <- 0
  df_aus[[paste0(item, "_n")]][df_aus[[item]] == "Some of the time"] <- 1
  df_aus[[paste0(item, "_n")]][df_aus[[item]] == "Most of the time"] <- 2
  df_aus[[paste0(item, "_n")]][df_aus[[item]] == "All or almost all of the time"] <- 3
}
df_aus$wrhpp_n = ifelse(!is.na(df_aus$wrhpp_n), 3 - df_aus$wrhpp_n, NA)
df_aus$enjlf_n = ifelse(!is.na(df_aus$enjlf_n), 3 - df_aus$enjlf_n, NA)
df_aus$CESD_TOTAL = rowSums(df_aus[, paste0(cesd_items, "_n")], na.rm = FALSE)

3 Reliability of the Depression Scale

alpha_value = alpha(df_aus[, paste0(cesd_items, "_n")])
alpha_value$total$raw_alpha

## [1] 0.8033533

The CES-D8 scale shows high internal consistency, with a Cronbach’s Alpha of r round(alpha_value\(total\)raw_alpha, 3). This indicates that all eight items reliably measure the same underlying construct—depression. None of the items substantially improves reliability when removed, suggesting that each item contributes meaningfully to the overall score.

4 Descriptive Statistics and Normality

summary(df_aus$CESD_TOTAL)

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##    0.00    2.00    4.00    4.87    7.00   24.00      33

min(df_aus$CESD_TOTAL, na.rm=TRUE)

## [1] 0

max(df_aus$CESD_TOTAL, na.rm=TRUE)

## [1] 24

mean(df_aus$CESD_TOTAL, na.rm=TRUE)

## [1] 4.869884

shapiro.test(df_aus$CESD_TOTAL)

## 
##  Shapiro-Wilk normality test
## 
## data:  df_aus$CESD_TOTAL
## W = 0.9133, p-value < 2.2e-16

The observed CES-D8 scores range from 0 to 24, with a mean of 4.87. This suggests that, on average, Austrian respondents report relatively low to moderate levels of depressive symptoms.

ggplot(df_aus[!is.na(df_aus$CESD_TOTAL), ], aes(x = CESD_TOTAL)) +
  geom_histogram(binwidth = 1, fill = "pink", color = "black") +
  labs(
    title = "Distribution of CES-D8 Depression Scores",
    subtitle = "ESS Round 11, Austria",
    x = "CES-D8 Score",
    y = "Frequency",
    caption = "Source: ESS11, own calculation"
  ) +
  theme_minimal()

The histogram shows that CES-D8 scores are right-skewed, with most values concentrated between 3 and 10. The Shapiro-Wilk test confirms that the distribution of depression scores significantly deviates from normality (p < 0.05). Therefore, the analysis proceeds with non-parametric methods where appropriate (e.g., Spearman’s rho and Kruskal-Wallis tests) to avoid violations of distributional assumptions.

5 Bivariate Analysis

# Recode: Income Satisfaction
df_aus$hincfel_num = NA  
df_aus$hincfel_num[df_aus$hincfel == "Living comfortably on present income"] = 3
df_aus$hincfel_num[df_aus$hincfel == "Coping on present income"] = 2
df_aus$hincfel_num[df_aus$hincfel == "Difficult on present income"] = 1
df_aus$hincfel_num[df_aus$hincfel == "Very difficult on present income"] = 0

# Recode: Health System Satisfaction
df_aus$stfhlth_num = as.numeric(factor(df_aus$stfhlth))

# Recode: Childhood Conflict
df_aus$cnfpplh_num = as.numeric(factor(
  df_aus$cnfpplh,
  levels = c("Always", "Often", "Sometimes", "Hardly ever", "Never"),
  ordered = TRUE
))

# Working Hours
df_aus$wkhtot = as.numeric(as.character(df_aus$wkhtot))

5.1 Income Satisfaction and Depression

cor.test(df_aus$CESD_TOTAL, df_aus$hincfel_num, method = "spearman")

## 
##  Spearman's rank correlation rho
## 
## data:  df_aus$CESD_TOTAL and df_aus$hincfel_num
## S = 2562360194, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
##       rho 
## -0.252128

kruskal.test(CESD_TOTAL ~ hincfel_num, data = df_aus)

## 
##  Kruskal-Wallis rank sum test
## 
## data:  CESD_TOTAL by hincfel_num
## Kruskal-Wallis chi-squared = 176.62, df = 3, p-value < 2.2e-16

ggplot(df_aus, aes(x = as.factor(hincfel_num), y = CESD_TOTAL)) +
  geom_boxplot(fill = "lightblue", color = "black") +
  labs(
    title = "Depressive Symptoms by Income Satisfaction",
    x = "Income Satisfaction (0 = Very difficult, 3 = Living comfortably)",
    y = "CES-D8 Score",
    caption = "ESS Round 11 – Austria"
  ) +
  theme_minimal()

The Spearman correlation confirms a statistically significant negative relationship between income satisfaction and depressive symptoms (ρ = -0.252).

The Kruskal-Wallis test further supports significant group differences across income categories (p < 0.001). The boxplot reveals that respondents facing greater financial difficulties report markedly higher CES-D8 scores, indicating stronger depressive symptoms. Conclusion: This finding supports Hypothesis 1 (H1) – higher income satisfaction is associated with lower depressive symptoms.

5.2 Health Services Satisfaction and Depression

cor.test(df_aus$CESD_TOTAL, df_aus$stfhlth_num, method = "spearman")

## 
##  Spearman's rank correlation rho
## 
## data:  df_aus$CESD_TOTAL and df_aus$stfhlth_num
## S = 2421778021, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
##        rho 
## -0.1712044

kruskal.test(CESD_TOTAL ~ stfhlth_num, data = df_aus)

## 
##  Kruskal-Wallis rank sum test
## 
## data:  CESD_TOTAL by stfhlth_num
## Kruskal-Wallis chi-squared = 76.065, df = 10, p-value = 2.951e-12

The correlation between confidence in health services and depression is statistically significant and negative. Individuals who express greater satisfaction with healthcare tend to report lower levels of depressive symptoms. The Kruskal-Wallis test also indicates significant group differences in CES-D8 scores across satisfaction levels. Conclusion: The results confirm Hypothesis 2 (H2) – positive perceptions of health services are linked to lower depression scores.

5.3 Childhood Conflicts and Depression

cor.test(df_aus$CESD_TOTAL, df_aus$cnfpplh_num, method = "spearman")

## 
##  Spearman's rank correlation rho
## 
## data:  df_aus$CESD_TOTAL and df_aus$cnfpplh_num
## S = 2608694583, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
##        rho 
## -0.2747699

kruskal.test(CESD_TOTAL ~ cnfpplh_num, data = df_aus)

## 
##  Kruskal-Wallis rank sum test
## 
## data:  CESD_TOTAL by cnfpplh_num
## Kruskal-Wallis chi-squared = 175.57, df = 4, p-value < 2.2e-16

The Spearman correlation shows a positive and significant association between the frequency of childhood conflicts and adult depression levels. Individuals who often experienced conflicts during childhood tend to report higher CES-D8 scores, suggesting lasting psychological effects. Conclusion: This finding confirms Hypothesis 3 (H3) – adverse childhood experiences are significantly associated with adult depressive symptoms.

5.4 Working Hours and Depression

cor.test(df_aus$CESD_TOTAL, df_aus$wkhtot, method = "spearman")

## 
##  Spearman's rank correlation rho
## 
## data:  df_aus$CESD_TOTAL and df_aus$wkhtot
## S = 1510067318, p-value = 0.5892
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
##       rho 
## 0.0118111

ggplot(df_aus, aes(x = wkhtot, y = CESD_TOTAL)) +
  geom_point(alpha = 0.5, color = "blue") +
  geom_smooth(method = "lm", color = "red", se = FALSE) +
  labs(
    title = "Depressive Symptoms vs. Weekly Working Hours",
    x = "Weekly Working Hours",
    y = "CES-D8 Score",
    caption = "ESS Round 11 – Austria"
  ) +
  theme_minimal()

The correlation between working hours and depressive symptoms is weak and not statistically significant (ρ ≈ 0, p > 0.05). The scatterplot shows no clear trend, and the fitted line remains mostly flat. This suggests that within this sample, working time alone is not a reliable predictor of depression. Conclusion: Hypothesis 4 (H4) is not supported – weekly working hours do not significantly affect depression scores.

6 Multiple Linear Regression

cor_matrix = cor(df_aus[, c("CESD_TOTAL", "hincfel_num", "cnfpplh_num", "wkhtot", "stfhlth_num")], 
                 use = "pairwise.complete.obs")
print(cor_matrix)

##              CESD_TOTAL hincfel_num cnfpplh_num       wkhtot  stfhlth_num
## CESD_TOTAL   1.00000000 -0.28143950  -0.2689294 -0.011487895 -0.167037293
## hincfel_num -0.28143950  1.00000000   0.1493801  0.041298772  0.047046449
## cnfpplh_num -0.26892936  0.14938008   1.0000000 -0.010915399  0.163197732
## wkhtot      -0.01148789  0.04129877  -0.0109154  1.000000000  0.001021815
## stfhlth_num -0.16703729  0.04704645   0.1631977  0.001021815  1.000000000

mlrm = lm(CESD_TOTAL ~ hincfel_num + cnfpplh_num + wkhtot + stfhlth_num, data = df_aus)
summary(mlrm)

## 
## Call:
## lm(formula = CESD_TOTAL ~ hincfel_num + cnfpplh_num + wkhtot + 
##     stfhlth_num, data = df_aus)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -7.9785 -2.1897 -0.6116  1.6286 18.3368 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 11.866608   0.473159  25.080  < 2e-16 ***
## hincfel_num -1.230848   0.103305 -11.915  < 2e-16 ***
## cnfpplh_num -0.723075   0.074471  -9.710  < 2e-16 ***
## wkhtot      -0.002867   0.006769  -0.424    0.672    
## stfhlth_num -0.183817   0.031195  -5.892 4.43e-09 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.29 on 2068 degrees of freedom
##   (281 observations deleted due to missingness)
## Multiple R-squared:  0.1386, Adjusted R-squared:  0.1369 
## F-statistic: 83.17 on 4 and 2068 DF,  p-value: < 2.2e-16

The multiple linear regression model predicts the CES-D8 depression score based on four social determinants:

Income satisfaction (hincfel_num)
Childhood conflict (cnfpplh_num)
Working hours (wkhtot)
Satisfaction with health services (stfhlth_num)

The model is statistically significant overall:

F(4, 2068) = 83.17,
p < <2e-16
The model explains 13.9% of the variance in CES-D8 scores (adjusted R² = 13.7%).

This indicates a moderate model fit, which is acceptable for social science research involving self-reported data.

💰 Income satisfaction

Each unit increase in income satisfaction (i.e., moving toward “living comfortably”) is associated with a -1.23 point decrease in depression scores (p < 0.001).

👶 Childhood conflict

For each step increase in reported childhood conflict frequency, depression scores increase by -0.72 points (p < 0.001).

🏥 Satisfaction with health services

Lower satisfaction with the health system is significantly associated with -0.18 points higher CES-D8 scores (p < 0.001).

🕒 Weekly working hours

Working hours have a non-significant effect on depression (-0.003 points per hour, p = 0.672).
This confirms previous results from the bivariate analysis.

✅ Summary: The regression analysis reinforces earlier findings: Income dissatisfaction, childhood adversity, and dissatisfaction with healthcare are robust and statistically significant predictors of higher depression levels.
In contrast, weekly working hours do not show a meaningful association with depressive symptoms in this sample.

7 Logistic Regression

df_aus$depression_binary = ifelse(df_aus$CESD_TOTAL >= 9, 1, 0)

table(df_aus$depression_binary)

## 
##    0    1 
## 1973  348

round(mean(df_aus$depression_binary == 1, na.rm = TRUE) * 100)

## [1] 15

Based on the clinical threshold (CES-D8 ≥ 9), 348 respondents, or approximately 15%, are classified as experiencing clinically significant depressive symptoms.

# Recode age
df_aus$agea = as.numeric(as.character(df_aus$agea))
df_aus$age_group = cut(df_aus$agea,
                       breaks = c(15, 24, 34, 44, 54, 64, Inf),
                       labels = c("15–24", "25–34", "35–44", "45–54", "55–64", "65+"),
                       right = TRUE,
                       include.lowest = TRUE)

# Recode general health into binary
df_aus$health_binary = ifelse(as.character(df_aus$health) %in% c("Very good", "Good"), 1, 0)

model_logit = glm(
  depression_binary ~ age_group + gndr + hincfel_num + cnfpplh_num + wkhtot + stfhlth_num + health_binary,
  data = df_aus, family = binomial
)
summary(model_logit)

## 
## Call:
## glm(formula = depression_binary ~ age_group + gndr + hincfel_num + 
##     cnfpplh_num + wkhtot + stfhlth_num + health_binary, family = binomial, 
##     data = df_aus)
## 
## Coefficients:
##                 Estimate Std. Error z value Pr(>|z|)    
## (Intercept)     2.899754   0.505327   5.738 9.56e-09 ***
## age_group25–34 -1.029654   0.383213  -2.687 0.007212 ** 
## age_group35–44 -0.534802   0.344712  -1.551 0.120795    
## age_group45–54 -0.937658   0.347189  -2.701 0.006919 ** 
## age_group55–64 -1.241529   0.343569  -3.614 0.000302 ***
## age_group65+   -0.720817   0.326242  -2.209 0.027143 *  
## gndrFemale      0.318084   0.145125   2.192 0.028394 *  
## hincfel_num    -0.459516   0.093464  -4.917 8.81e-07 ***
## cnfpplh_num    -0.362048   0.065960  -5.489 4.04e-08 ***
## wkhtot         -0.006567   0.006213  -1.057 0.290560    
## stfhlth_num    -0.076725   0.027929  -2.747 0.006012 ** 
## health_binary  -1.548431   0.153265 -10.103  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 1742.3  on 2072  degrees of freedom
## Residual deviance: 1453.7  on 2061  degrees of freedom
##   (281 observations deleted due to missingness)
## AIC: 1477.7
## 
## Number of Fisher Scoring iterations: 5

# Odds Ratios
exp(coef(model_logit))

##    (Intercept) age_group25–34 age_group35–44 age_group45–54 age_group55–64 
##     18.1696749      0.3571304      0.5857850      0.3915437      0.2889420 
##   age_group65+     gndrFemale    hincfel_num    cnfpplh_num         wkhtot 
##      0.4863548      1.3744916      0.6315895      0.6962491      0.9934545 
##    stfhlth_num  health_binary 
##      0.9261445      0.2125812

# 95% Confidence Intervals
exp(confint(model_logit))

##                    2.5 %     97.5 %
## (Intercept)    6.7177602 48.8688385
## age_group25–34 0.1684137  0.7626168
## age_group35–44 0.3017493  1.1726972
## age_group45–54 0.2005879  0.7871821
## age_group55–64 0.1491852  0.5770728
## age_group65+   0.2610072  0.9431630
## gndrFemale     1.0358754  1.8306165
## hincfel_num    0.5256087  0.7584013
## cnfpplh_num    0.6116095  0.7922323
## wkhtot         0.9813397  1.0055790
## stfhlth_num    0.8769269  0.9784701
## health_binary  0.1569675  0.2863900

The logistic regression model estimates the probability of being clinically depressed (CES-D8 ≥ 9) based on age group, gender, income satisfaction, childhood conflict, working hours, healthcare satisfaction, and general health.

🔹 Age Group

Compared to the youngest group (15–24), the odds of clinical depression tend to decrease in older age groups. However, this trend is not statistically significant across all groups, suggesting that age alone is not a strong predictor in this sample.

🔹 Gender

Women appear to be at slightly higher risk of clinical depression, but the association is not statistically significant in the full model.

🔹 Income Satisfaction

Higher income satisfaction is significantly associated with lower odds of clinical depression. For each step toward greater financial ease, the odds of being depressed decrease markedly.

🔹 Childhood Conflict

More frequent childhood conflict experiences lead to higher odds of adult depression. This effect remains significant even when controlling for other factors, confirming the lasting impact of early adversity.

🔹 Satisfaction with Health Services

Dissatisfaction with healthcare services is linked to a significant increase in the odds of clinical depression. This confirms findings from the linear model.

🔹 Self-Rated Health

The strongest effect comes from general health status: individuals who rate their health as fair or poor have much higher odds of depression. This association is highly significant and aligns with clinical expectations.

🔹 Working Hours

As in the linear model, working hours show no significant association with depression risk.

r_mcfadden <- 1 - (model_logit$deviance / model_logit$null.deviance)
n <- nobs(model_logit)
r_nagelkerke <- r_mcfadden / (1 - (model_logit$null.deviance / (n * log(2))))

The model explains approximately 16.6% (McFadden’s R²) of the variation in the outcome. The Nagelkerke R² is -77.9%, indicating a meaningful effect size for behavioral data.

Summary

The logistic regression confirms key findings from the linear model: income dissatisfaction, childhood conflict, and perceived poor health services are strong predictors of depression. Additionally, self-rated health shows the largest effect on clinical depression, highlighting the relevance of general wellbeing as a protective or risk factor. Working hours and gender, by contrast, do not significantly contribute to the prediction of clinical depression in this dataset.

8 Extended Hypothesis: Interaction Between Income Satisfaction and Gender

To explore whether the relationship between income satisfaction and depression differs between men and women, we extend our model with an interaction term: hincfel_num * gndr. This allows us to test the hypothesis:

H5: The protective effect of higher income satisfaction on depression is stronger for women than for men.

# Recode gender as factor
df_aus$gndr <- as.factor(df_aus$gndr)

# Interaction model
model_interact <- glm(
  depression_binary ~ hincfel_num * gndr + cnfpplh_num + wkhtot + stfhlth_num + health_binary,
  data = df_aus,
  family = binomial
)

The table below summarizes the model results using tab_model() for clearer interpretation of odds ratios and significance levels:

tab_model(model_interact,
          show.ci = TRUE,
          show.se = TRUE,
          show.p = TRUE,
          digits = 3,
          title = "Logistic Regression with Interaction: Income Satisfaction × Gender")

Logistic Regression with Interaction: Income Satisfaction × Gender
	depression_binary
Predictors	Odds Ratios	std. Error	CI	p
(Intercept)	8.737	4.362	0.000 – Inf	<0.001
hincfel num	0.597	0.091	0.000 – Inf	0.001
gndr [Female]	1.109	0.444	0.000 – Inf	0.797
cnfpplh num	0.698	0.045	0.000 – Inf	<0.001
wkhtot	0.993	0.006	0.000 – Inf	0.253
stfhlth num	0.928	0.026	0.000 – Inf	0.007
health binary	0.223	0.030	0.000 – Inf	<0.001
hincfel num × gndr [Female]	1.112	0.211	0.000 – Inf	0.575
Observations	2073
R² Tjur	0.153

To better understand the interaction, we visualize predicted probabilities across levels of income satisfaction for both genders:

plot_model(model_interact, type = "int", terms = c("hincfel_num", "gndr")) +
  labs(
    title = "Interaction Effect: Income Satisfaction × Gender",
    y = "Predicted Probability of Depression",
    x = "Income Satisfaction"
  )

To further explore the role of income satisfaction in predicting depression risk, we estimated a logistic regression model including an interaction term between income satisfaction and gender. The results show that income satisfaction remains a significant protective factor
(OR = 0.597, p = 0.001), confirming earlier findings.

However, the interaction term (income satisfaction × gender) is not statistically significant
(OR = 1.112, p = 0.575), indicating that the strength of the income effect does not differ significantly between men and women.

Similarly, the main effect of gender is non-significant (p = 0.797), although women show slightly elevated odds of depression across income levels.

The predicted probabilities visualized above illustrate this pattern: Both genders benefit similarly from higher income satisfaction, but women consistently exhibit higher predicted probabilities of depression, regardless of income status.

These results suggest no strong gender moderation, but highlight persistent gender disparities in baseline mental health risk. Future research could examine whether these disparities stem from gendered experiences of stress, caregiving, or access to resources.

9 Graphics

9.1 Depressive Symptoms by Binary Income Satisfaction

df_aus$hincfel_bin <- ifelse(df_aus$hincfel_num %in% c(2, 3), 1,  # 2 = coping, 3 = comfortable
                             ifelse(df_aus$hincfel_num %in% c(0, 1), 0, NA))

ggplot(df_aus[!is.na(df_aus$CESD_TOTAL) & !is.na(df_aus$hincfel_bin), ],
       aes(x = factor(hincfel_bin), y = CESD_TOTAL)) +
  geom_boxplot(fill = "steelblue") +
  labs(
    title = "Depressive Symptoms by Binary Income Satisfaction",
    subtitle = "ESS Round 11 – Austria",
    x = "Income Satisfaction (0 = Difficulty, 1 = Coping/Comfortable)",
    y = "CES-D8 Depression Score",
    caption = "Krabichler & Wachter (2025)"
  ) +
  theme_minimal()

The boxplot clearly shows that individuals experiencing financial difficulty report higher levels of depressive symptoms than those who are coping or living comfortably. The median CES-D8 score is notably elevated in the “difficulty” group, and the interquartile range indicates a broader distribution of depressive experiences under financial strain. These results visually reinforce the earlier regression findings and provide further evidence that income-related stress is a major risk factor for mental health.

9.2 Clinically Significant Depression by Self-Rated Health

df_plot <- droplevels(df_aus[complete.cases(df_aus[, c("health", "depression_binary")]), ])

ggplot(df_plot, aes(x = health)) +
  geom_bar(aes(fill = factor(depression_binary)), position = "fill", width = 0.6) +
  scale_y_continuous(labels = scales::percent) +
  scale_fill_manual(values = c("steelblue", "red"), labels = c("No Depression", "Depressed")) +
  labs(
    title = "Clinically Significant Depression by Self-Rated Health",
    subtitle = "ESS Round 11 – Austria",
    x = "Self-Rated Health",
    y = "Share of respondents",
    fill = "Depression Status",
    caption = "Krabichler & Wachter (2025)"
  ) +
  theme_minimal()

The chart displays a clear gradient: as self-rated health worsens, the share of respondents with clinical depression (CES-D8 ≥ 9) increases sharply. Among those rating their health as “very good”, fewer than 10% are depressed, compared to nearly 70% of those reporting “very bad” health. This pattern confirms that subjective health status is a powerful indicator of depression risk and supports the inclusion of this variable in predictive models.

9.3 Self-Rated Health by Gender

ggplot(df_aus[!is.na(df_aus$health) & !is.na(df_aus$gndr), ], aes(x = gndr)) +
  geom_bar(aes(fill = health), position = "fill", width = 0.6) +
  scale_y_continuous(labels = scales::percent) +
  scale_fill_manual(values = c("darkgreen", "lightgreen", "grey", "orange", "red")) +
  coord_flip() +
  labs(
    title = "Self-Rated Health by Gender",
    subtitle = "ESS Round 11 – Austria",
    x = "Gender",
    y = "Percentage Distribution",
    fill = "Health Status",
    caption = "Krabichler & Wachter (2025)"
  ) +
  theme_minimal()

Women report worse subjective health than men. A higher proportion of women rate their health as “fair” or “poor”, while men more frequently select “very good” or “good”. These differences may reflect both real disparities in health conditions and gendered patterns of health perception or reporting. As women also showed a (non-significant) trend toward higher depression risk, this could point to indirect effects through perceived health.

10 Final Remarks

This study aimed to examine how social determinants—particularly income satisfaction, childhood experiences, perceived health services, and working hours—are related to depressive symptoms in Austria, using data from the European Social Survey (ESS11).

The findings show that a significant share of the population exceeds the clinical threshold for depression (CES-D8 ≥ 9), with approximately 15% of respondents being classified as clinically depressed. The average depression score across the population was 4.87, and the distribution of scores was right-skewed, suggesting that severe depressive symptoms affect a smaller—but substantial—subgroup.

10.1 Key Findings

Both bivariate and multivariate analyses confirm that social inequality and personal adversity play a key role in shaping mental health outcomes:

Income satisfaction was consistently and significantly associated with lower levels of depression. Respondents who reported financial difficulty exhibited notably higher CES-D8 scores and greater odds of clinical depression.
Childhood conflict also emerged as a powerful predictor. Even when controlling for other factors, individuals who experienced more frequent conflicts during childhood were significantly more likely to suffer from depressive symptoms in adulthood.
Satisfaction with the healthcare system was negatively associated with depression. This suggests that institutional trust and perceived quality of services may influence mental wellbeing.
Self-rated general health was the strongest predictor in the logistic model. Those who rated their health as poor were several times more likely to be clinically depressed. This relationship underscores the importance of holistic health perspectives that consider both physical and mental dimensions.
By contrast, working hours showed no significant relationship with depression. This finding suggests that the quantity of work alone may not be as relevant as the quality of work, job security, or work-life balance.

10.2 Theoretical and Practical Implications

These results contribute to the growing literature on the social determinants of mental health, confirming that both early-life experiences and current socio-economic conditions have lasting effects on psychological wellbeing. The consistent role of income satisfaction and subjective health evaluations highlights the need for policies that reduce financial stress, promote preventive mental health care, and ensure accessible, trustworthy health services.

From a public health perspective, interventions should target vulnerable groups, especially individuals reporting low income, poor health, or adverse childhood experiences. Improving perceptions of public institutions such as the healthcare system may indirectly promote better mental health outcomes.

10.3 Reflection

The findings also open new directions for further research. Gender differences in health perception, potential mediating mechanisms (e.g., health satisfaction between self-rated health and depression), and social context factors could be explored through advanced methods such as mediation, moderation, or multilevel modeling.

Furthermore, while the CES-D8 is a widely used scale, it remains self-reported. Future studies could complement survey data with clinical assessments or longitudinal data to better understand the dynamics of depression across time.

Social Determinants of Depression in Austria

Chiara Wachter & Elias Krabichler

2025-07-05