Homework: Association Rules
Natalia Szymanko
Data Preparation
To investigate the factors influencing a household’s financial well-being, a broad set of socio-economic and value-based indicators was selected from the ESS database.
## 'data.frame': 50116 obs. of 25 variables:
## $ name : chr "ESS11e04_1" "ESS11e04_1" "ESS11e04_1" "ESS11e04_1" ...
## $ essround: int 11 11 11 11 11 11 11 11 11 11 ...
## $ edition : num 4.1 4.1 4.1 4.1 4.1 4.1 4.1 4.1 4.1 4.1 ...
## $ proddate: chr "12.01.2026" "12.01.2026" "12.01.2026" "12.01.2026" ...
## $ idno : int 50014 50030 50057 50106 50145 50158 50211 50212 50213 50235 ...
## $ cntry : chr "AT" "AT" "AT" "AT" ...
## $ dweight : num 1.185 0.61 1.392 0.556 0.723 ...
## $ pspwght : num 0.393 0.325 4 0.176 1.061 ...
## $ pweight : num 0.331 0.331 0.331 0.331 0.331 ...
## $ anweight: num 0.13 0.1076 1.3237 0.0583 0.3511 ...
## $ lrscale : int 5 0 3 5 2 4 4 3 5 5 ...
## $ stflife : int 8 9 10 7 9 8 8 8 8 8 ...
## $ stfgov : int 4 5 5 4 7 2 3 6 5 8 ...
## $ gincdif : int 2 1 1 1 2 2 2 2 1 2 ...
## $ health : int 3 2 1 3 2 1 2 2 2 2 ...
## $ rlgdgr : int 5 0 8 6 1 3 6 6 5 10 ...
## $ brncntr : int 1 1 1 2 1 2 1 1 1 1 ...
## $ eisced : int 3 5 6 5 3 4 2 7 5 2 ...
## $ uemp3m : int 2 2 2 2 2 2 1 2 2 2 ...
## $ hincfel : int 1 2 1 2 2 1 2 1 2 2 ...
## $ impricha: int 5 4 4 4 4 2 4 6 3 4 ...
## $ iphlppla: int 2 1 1 2 2 1 2 2 1 3 ...
## $ prob : num 0.000579 0.001124 0.000493 0.001233 0.000949 ...
## $ stratum : int 107 69 18 101 115 7 58 38 62 105 ...
## $ psu : int 317 128 418 295 344 373 86 3 108 314 ...Using the ESS Round 11 Codebook, we identified and removed non-substantive responses. We treated values such as “Refusal” (7/77), “Don’t know” (8/88), and “No answer” (9/99) as NA and excluded them from the analysis to ensure the integrity of the patterns discovered by the algorithm. The variables were also given clear names.
variables <- c("hincfel", "lrscale", "stflife", "stfgov", "gincdif",
"health", "rlgdgr", "brncntr", "eisced", "uemp3m",
"impricha", "iphlppla")
data <- data[, variables]
short_scales <- c("hincfel", "health", "brncntr", "uemp3m", "impricha", "iphlppla", "gincdif")
for(var in short_scales) {
data[[var]][data[[var]] > 6] <- NA
}
long_scales <- c("lrscale", "stflife", "stfgov", "rlgdgr", "eisced")
for(var in long_scales) {
data[[var]][data[[var]] > 10] <- NA
}
data <- na.omit(data)
data <- data %>% rename(
income_feeling = hincfel,
left_right = lrscale,
life_sat = stflife,
gov_sat = stfgov,
redistribution = gincdif,
subjective_health = health,
religiosity = rlgdgr,
born_in_country = brncntr,
education_level = eisced,
unemployed_3m = uemp3m,
importance_rich = impricha,
importance_helping = iphlppla
)Since the Apriori algorithm requires categorical data (factors) rather than continuous numbers, we transformed all numerical scales into meaningful groups. Variables with 0–10 Scales were grouped into three tiers: Low (0–3), Medium (4–6), and High (7–10). Health and Values were consolidated into intuitive categories like “Good/Fair/Poor” or “High/Medium/Low” based on their original coding logic. Binary Indicators were transformed into “Yes/No” factors (for unemployment and birth country).
data_final <- data %>%
mutate(
income_feeling = factor(income_feeling,
levels = c(1, 2, 3, 4),
labels = c("Comfortable", "Coping", "Difficult", "Very_Difficult")),
ideology = cut(left_right, breaks = c(-1, 3, 6, 10), labels = c("Left", "Center", "Right")),
life_sat_cat = cut(life_sat, breaks = c(-1, 3, 6, 10), labels = c("Low", "Medium", "High")),
gov_sat_cat = cut(gov_sat, breaks = c(-1, 3, 6, 10), labels = c("Low", "Medium", "High")),
religiosity_cat = cut(religiosity, breaks = c(-1, 3, 6, 10), labels = c("Low", "Medium", "High")),
health_cat = cut(subjective_health, breaks = c(0, 2, 3, 5), labels = c("Good", "Fair", "Poor")),
redist_opinon = cut(redistribution, breaks = c(0, 2, 3, 5), labels = c("Agree", "Neutral", "Disagree")),
born_here = factor(born_in_country, levels = c(1, 2), labels = c("Yes", "No")),
unemployed = factor(unemployed_3m, levels = c(1, 2), labels = c("Yes", "No")),
edu_cat = cut(education_level, breaks = c(0, 2, 5, 8), labels = c("Low", "Medium", "High")),
imp_rich = cut(importance_rich, breaks = c(0, 2, 4, 6), labels = c("High", "Medium", "Low")),
imp_helping = cut(importance_helping, breaks = c(0, 2, 4, 6), labels = c("High", "Medium", "Low"))
) %>%
select_if(is.factor)| Variable | Description | Categories |
|---|---|---|
| income_feeling | Feeling about household’s income nowadays | Comfortable, Coping, Difficult, Very_Difficult |
| ideology | Placement on left-right scale | Left (0-3), Center (4-6), Right (7-10) |
| life_sat_cat | How satisfied with life as a whole | Low (0-3), Medium (4-6), High (7-10) |
| gov_sat_cat | How satisfied with the national government | Low (0-3), Medium (4-6), High (7-10) |
| redist_opinion | Government should reduce income differences | Agree, Neutral, Disagree |
| health_cat | Subjective general health | Good, Fair, Poor |
| religiosity_cat | How religious are you | Low, Medium, High |
| born_here | Born in the country | Yes, No |
| edu_cat | Highest level of education (EISCED) | Low, Medium, High |
| unemployed | Ever unemployed and seeking work for 3 months | Yes, No |
| imp_rich | Important to be rich, have money and expensive things | High, Medium, Low |
| imp_helping | Important to help people and care for others’ well-being | High, Medium, Low |
Then, the data frame is converted into a matrix of transactions.
Exploratory Analysis
The summary() output reveals the fundamental structure of our data. Density equal to 0.343 indicates that roughly 34% of the cells in our transaction matrix contain a value. Since every respondent has exactly 12 attributes (length distribution = 12), the matrix is balanced. The most common traits in the population are being born in the country (born_here=Yes), not having a history of unemployment (unemployed=No), and agreeing with income redistribution (redist_opinon=Agree). High life satisfaction and high altruism (imp_helping=High) are also dominant characteristics in this European sample.
## transactions as itemMatrix in sparse format with
## 39821 rows (elements/itemsets/transactions) and
## 35 columns (items) and a density of 0.3428571
##
## most frequent items:
## born_here=Yes unemployed=No redist_opinon=Agree life_sat_cat=High
## 36522 29347 28950 27849
## imp_helping=High (Other)
## 27663 327521
##
## element (itemset/transaction) length distribution:
## sizes
## 12
## 39821
##
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 12 12 12 12 12 12
##
## includes extended item information - examples:
## labels variables levels
## 1 income_feeling=Comfortable income_feeling Comfortable
## 2 income_feeling=Coping income_feeling Coping
## 3 income_feeling=Difficult income_feeling Difficult
##
## includes extended transaction information - examples:
## transactionID
## 1 1
## 2 2
## 3 3Using the itemFrequency function, we analyzed the distribution of our target variable, Income Feeling. It shows that 34.2% of respondents feel “Comfortable” with their income, while the majority (44.7%) are “Coping”. A combined 21% of the population faces financial hardship, with 15.9% finding it “Difficult” and 5.2% “Very Difficult”. Because the “Very Difficult” category is relatively rare (low support), we will need to set a low supp threshold in the Apriori algorithm to find rules for this specific group.
## income_feeling=Comfortable income_feeling=Coping
## 0.34228171 0.44710078
## income_feeling=Difficult income_feeling=Very_Difficult
## 0.15906180 0.05155571Item Frequency Plot displays the 15 most frequent items in the dataset. It confirms that the dataset is dominated by stable socio-economic indicators (born here, employed, high life satisfaction).
itemFrequencyPlot(data_t, topN = 15, col=moj_kolor, type = "relative", main = "Item Frequency Plot")
Next, we generated an image() plot for a random sample of 100 transactions. Each dot represents the presence of an item in a transaction. The vertical columns of dots indicate very common items (high support). The lack of obvious horizontal blocks suggests that there isn’t one single “type” of person that represents everyone, which is ideal for discovering diverse association rules through the Apriori algorithm.
Association Rule Learning
To extract meaningful associations, the Apriori algorithm was configured with a minimum support of 0.001 and confidence of 0.20. The low support threshold was intentionally selected to capture patterns related to the ‘Very Difficult’ income category, which represents a small minority of the sample. The Right-Hand Side was restricted to the income_feeling levels to ensure that all discovered rules provide direct insights into the socio-economic drivers of subjective financial well-being.
rules <- apriori(data = data_t,
parameter = list(supp = 0.001, conf = 0.2, minlen = 2),
appearance = list(default = "lhs",
rhs = c("income_feeling=Comfortable",
"income_feeling=Coping",
"income_feeling=Difficult",
"income_feeling=Very_Difficult")),
control = list(verbose = F))First, we focus on rules with the highest lift. This analysis reveals a powerful link between subjective financial distress and a cluster of negative socio-economic indicators. The strongest rule, with a Lift of 9.47, identifies individuals characterized by low life satisfaction, poor health, and a history of unemployment as being significantly more likely to report ‘Very Difficult’ living conditions. Interestingly, while these individuals represent a small segment of the population (Support around 0.1%), the Confidence of nearly 50% suggests that this specific profile is a high-accuracy predictor of extreme financial strain within the ESS dataset.
## lhs rhs support confidence coverage lift count
## [1] {ideology=Center,
## life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## born_here=Yes,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001054720 0.4883721 0.002159664 9.472706 42
## [2] {ideology=Center,
## life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001079832 0.4777778 0.002260114 9.267213 43
## [3] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## born_here=Yes,
## unemployed=Yes,
## imp_rich=Low} => {income_feeling=Very_Difficult} 0.001029607 0.4659091 0.002209889 9.037002 41
## [4] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## redist_opinon=Agree,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001732754 0.4630872 0.003741744 8.982269 69
## [5] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## redist_opinon=Agree,
## born_here=Yes,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001582080 0.4565217 0.003465508 8.854921 63
## [6] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## unemployed=Yes,
## edu_cat=Medium} => {income_feeling=Very_Difficult} 0.001104945 0.4536082 0.002435901 8.798409 44The results sorted by Support predominantly feature single-item antecedents. This is a direct consequence of the Apriori algorithm: as more conditions are added to a rule, its frequency in the dataset, and thus its Support, inevitably decreases. The rules with the highest support represent the most generalized demographic baselines in the European sample. For instance, the association between being born in the country (born_here=Yes) and the ‘Coping’ income status appears in over 41% of all transactions. While these rules are simple, they provide the necessary context for the study, establishing that ‘Coping’ is the normative financial experience for the majority of the population. However, the Lift values remain near 1.0, suggesting these are general demographic baselines rather than specific predictors of financial situation.
## lhs rhs support confidence coverage lift count
## [1] {born_here=Yes} => {income_feeling=Coping} 0.4123452 0.4495920 0.9171543 1.0055720 16420
## [2] {redist_opinon=Agree} => {income_feeling=Coping} 0.3321614 0.4568912 0.7270033 1.0218976 13227
## [3] {unemployed=No} => {income_feeling=Coping} 0.3284448 0.4456674 0.7369730 0.9967940 13079
## [4] {born_here=Yes} => {income_feeling=Comfortable} 0.3129002 0.3411642 0.9171543 0.9967352 12460
## [5] {life_sat_cat=High} => {income_feeling=Coping} 0.3124482 0.4467665 0.6993546 0.9992523 12442
## [6] {redist_opinon=Agree,
## born_here=Yes} => {income_feeling=Coping} 0.3072248 0.4594067 0.6687426 1.0275238 12234An analysis of the rules sorted by Confidence identifies the profile of respondents with the highest financial security. The premier rule shows a 90.9% probability (Confidence = 0.909) that individuals with high life and government satisfaction, centrist views, and opposition to income redistribution will report living ‘comfortably’. This cluster of attributes acts as a near-perfect predictor of economic stability. Furthermore, the presence of high education and a strong emphasis on helping others within these high-confidence rules suggests that the ‘Comfortable’ segment consists largely of highly educated individuals who value social help but prefer personal agency over state-led redistribution.
## lhs rhs support confidence coverage lift count
## [1] {ideology=Center,
## life_sat_cat=High,
## gov_sat_cat=High,
## religiosity_cat=Low,
## redist_opinon=Disagree,
## unemployed=No,
## imp_rich=Medium} => {income_feeling=Comfortable} 0.001004495 0.9090909 0.001104945 2.655973 40
## [2] {life_sat_cat=High,
## gov_sat_cat=High,
## religiosity_cat=Low,
## redist_opinon=Disagree,
## unemployed=No,
## edu_cat=High,
## imp_rich=Medium,
## imp_helping=High} => {income_feeling=Comfortable} 0.001004495 0.8888889 0.001130057 2.596951 40
## [3] {ideology=Center,
## life_sat_cat=High,
## gov_sat_cat=High,
## religiosity_cat=Low,
## health_cat=Good,
## redist_opinon=Disagree,
## unemployed=No,
## imp_helping=High} => {income_feeling=Comfortable} 0.001180282 0.8867925 0.001330956 2.590826 47
## [4] {ideology=Center,
## life_sat_cat=High,
## gov_sat_cat=High,
## religiosity_cat=Low,
## redist_opinon=Disagree,
## unemployed=No,
## imp_helping=High} => {income_feeling=Comfortable} 0.001305844 0.8813559 0.001481630 2.574943 52
## [5] {ideology=Center,
## life_sat_cat=High,
## religiosity_cat=Low,
## health_cat=Good,
## redist_opinon=Disagree,
## unemployed=No,
## edu_cat=High,
## imp_rich=Medium,
## imp_helping=High} => {income_feeling=Comfortable} 0.001305844 0.8813559 0.001481630 2.574943 52
## [6] {ideology=Center,
## gov_sat_cat=High,
## religiosity_cat=Low,
## redist_opinon=Disagree,
## unemployed=No,
## imp_rich=Medium} => {income_feeling=Comfortable} 0.001029607 0.8723404 0.001180282 2.548604 41The scatter plot visualizes the distribution of 288,766 generated rules based on their Support (x-axis), Confidence (y-axis), and Lift (shading). The vast majority of rules are concentrated at very low support levels (near 0.0), indicating that most complex socio-economic patterns are specific to small sub-groups of the population. We observe a clear trade-off where rules with the highest support (reaching up to 0.40) tend to have moderate confidence levels around 0.4 to 0.5. The most significant rules, indicated by the lightest shading, are found in the low-to-moderate confidence and low-support region. These rules possess Lift values exceeding 7.5.
plot(rules, method = "scatterplot", col=moj_kolor, measure = c("support", "confidence"), shading = "lift")
Targeted Rules for “Very Difficult” Income
To specifically investigate the socio-economic drivers of extreme financial hardship, the Apriori algorithm was constrained to generate rules where the Right-Hand Side (RHS) was exclusively income_feeling=Very_Difficult.
Due to the relative rarity of this category in the ESS sample (5.2%), we applied a minimum support threshold of 0.001 and a confidence of 0.15. This approach prioritizes rules with a high Lift, identifying the unique risk clusters, such as the combination of long-term unemployment and poor health, that dramatically increase the probability of severe financial strain compared to the general population.
The discovered rules exhibit exceptionally high Lift values, ranging from 8.64 to 9.47. This indicates that individuals matching the profiles on the Left-Hand Side (LHS) are approximately nine times more likely to experience extreme financial hardship compared to the average European respondent. The results reveal a consistent cluster where multiple negative life factors overlap. Notably, every single top-10 rule includes both unemployed = Yes and health_cat = Poor. This confirms that the intersection of physical health issues and labor market exclusion is the most important driver of subjective poverty in the ESS Round 11 data. Rules 4 and 5 highlight that individuals in this state of distress almost universally support government intervention to reduce income inequality (redist_opinon = Agree). The constant presence of gov_sat_cat = Low across the rules suggests that financial strain is deeply connected to a lack of faith in national institutions, likely due to a perceived failure of the social safety net. While these “risk profiles” have low Support (appearing in roughly 0.1% of all transactions), their high Confidence and Lift make them invaluable for policy considerations. They prove that financial difficulty is not just about income levels, but a complex interaction of health, employment status, and psychological well-being.
rules_very_difficult <- apriori(data = data_t,
parameter = list(supp = 0.001, conf = 0.15, minlen = 2),
appearance = list(default = "lhs",
rhs = "income_feeling=Very_Difficult"),
control = list(verbose = F))
inspect(head(sort(rules_very_difficult, by = "lift"), 10))## lhs rhs support confidence coverage lift count
## [1] {ideology=Center,
## life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## born_here=Yes,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001054720 0.4883721 0.002159664 9.472706 42
## [2] {ideology=Center,
## life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001079832 0.4777778 0.002260114 9.267213 43
## [3] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## born_here=Yes,
## unemployed=Yes,
## imp_rich=Low} => {income_feeling=Very_Difficult} 0.001029607 0.4659091 0.002209889 9.037002 41
## [4] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## redist_opinon=Agree,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001732754 0.4630872 0.003741744 8.982269 69
## [5] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## redist_opinon=Agree,
## born_here=Yes,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001582080 0.4565217 0.003465508 8.854921 63
## [6] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## unemployed=Yes,
## edu_cat=Medium} => {income_feeling=Very_Difficult} 0.001104945 0.4536082 0.002435901 8.798409 44
## [7] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001958765 0.4534884 0.004319329 8.796084 78
## [8] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## unemployed=Yes,
## imp_rich=Low} => {income_feeling=Very_Difficult} 0.001079832 0.4526316 0.002385676 8.779465 43
## [9] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## born_here=Yes,
## unemployed=Yes} => {income_feeling=Very_Difficult} 0.001808091 0.4500000 0.004017980 8.728422 72
## [10] {life_sat_cat=Low,
## gov_sat_cat=Low,
## health_cat=Poor,
## born_here=Yes,
## unemployed=Yes,
## edu_cat=Medium} => {income_feeling=Very_Difficult} 0.001029607 0.4456522 0.002310339 8.644089 41plot(head(sort(rules_very_difficult, by="lift"), 15), method="paracoord", control=list(reorder=TRUE))
Targeted Rules for “Difficult” Income
For the second-to-lowest income category, “Difficult”, we follow the same analytical procedure as before. While the “Very Difficult” category represents extreme poverty, the “Difficult” category captures a broader segment of the population struggling with financial instability. Since this category has a higher baseline frequency (15.9%) than “Very Difficult” (5.2%), the Support and Confidence thresholds are set slightly higher (supp=0.005 and conf=0.25) to filter out weak associations.
The targeted analysis for this income category highlights a distinct socio-economic profile focused on educational barriers and institutional reliance. With Lift values exceeding 2.35, the results indicate that individuals with low education (edu_cat=Low) and moderate life satisfaction are significantly more likely to face chronic financial strain. A critical insight from these rules is that several high-confidence rules involve individuals who are currently employed (unemployed=No) yet still report financial difficulty. This suggests that for this segment, employment alone is insufficient to guarantee financial security, particularly when combined with low institutional trust (gov_sat_cat=Low) and a high demand for state-led wealth redistribution.
rules_difficult <- apriori(data = data_t,
parameter = list(supp = 0.005, conf = 0.25, minlen = 2),
appearance = list(default = "lhs",
rhs = "income_feeling=Difficult"),
control = list(verbose = F))
inspect(head(sort(rules_difficult, by = "lift"), 10))## lhs rhs support confidence coverage lift count
## [1] {life_sat_cat=Low,
## redist_opinon=Agree,
## edu_cat=Low} => {income_feeling=Difficult} 0.005148037 0.4035433 0.01275709 2.537022 205
## [2] {life_sat_cat=Medium,
## gov_sat_cat=Low,
## redist_opinon=Agree,
## edu_cat=Low,
## imp_helping=High} => {income_feeling=Difficult} 0.005173150 0.3872180 0.01335979 2.434387 206
## [3] {life_sat_cat=Medium,
## religiosity_cat=High,
## redist_opinon=Agree,
## unemployed=No,
## edu_cat=Low} => {income_feeling=Difficult} 0.005198262 0.3833333 0.01356068 2.409965 207
## [4] {life_sat_cat=Medium,
## gov_sat_cat=Low,
## redist_opinon=Agree,
## edu_cat=Low} => {income_feeling=Difficult} 0.008035961 0.3818616 0.02104417 2.400712 320
## [5] {life_sat_cat=Medium,
## gov_sat_cat=Low,
## redist_opinon=Agree,
## born_here=Yes,
## unemployed=No,
## edu_cat=Low} => {income_feeling=Difficult} 0.005399161 0.3791887 0.01423872 2.383908 215
## [6] {life_sat_cat=Medium,
## gov_sat_cat=Low,
## redist_opinon=Agree,
## unemployed=No,
## edu_cat=Low} => {income_feeling=Difficult} 0.005600060 0.3786078 0.01479119 2.380256 223
## [7] {life_sat_cat=Medium,
## gov_sat_cat=Low,
## redist_opinon=Agree,
## born_here=Yes,
## edu_cat=Low} => {income_feeling=Difficult} 0.007533713 0.3773585 0.01996434 2.372402 300
## [8] {life_sat_cat=Medium,
## gov_sat_cat=Low,
## religiosity_cat=High,
## redist_opinon=Agree,
## born_here=Yes,
## unemployed=No} => {income_feeling=Difficult} 0.006428769 0.3748170 0.01715175 2.356424 256
## [9] {life_sat_cat=Medium,
## gov_sat_cat=Low,
## religiosity_cat=High,
## redist_opinon=Agree,
## born_here=Yes,
## imp_helping=High} => {income_feeling=Difficult} 0.006604555 0.3746439 0.01762889 2.355335 263
## [10] {life_sat_cat=Low,
## edu_cat=Low} => {income_feeling=Difficult} 0.006177645 0.3744292 0.01649883 2.353986 246
Targeted Rules for “Coping” Income
The targeted analysis for the ‘Coping’ category reveals the socio-economic standard of the European population. With Support values reaching nearly 7.3% for complex rules, this group represents the most common demographic profile in the dataset.
The findings indicate that the ‘Coping’ segment is primarily composed of individuals with medium education and a history of stable employment (unemployed=No). Interestingly, high life satisfaction is frequently associated with this category, suggesting that financial ‘adequacy’, rather than wealth, is sufficient for a positive life outlook for a large portion of respondents.
The Lift values near 1.2 confirm that this is a baseline group: these traits are widely distributed across the sample and represent the normative social experience where individuals manage their finances without significant distress but also without the luxury of the ‘Comfortable’ class.”
rules_coping <- apriori(data = data_t,
parameter = list(supp = 0.05, conf = 0.45, minlen = 2),
appearance = list(default = "lhs",
rhs = "income_feeling=Coping"),
control = list(verbose = F))
inspect(head(sort(rules_coping, by = "lift"), 10))## lhs rhs support confidence coverage lift count
## [1] {ideology=Center,
## life_sat_cat=High,
## redist_opinon=Agree,
## born_here=Yes,
## edu_cat=Medium} => {income_feeling=Coping} 0.06858190 0.5323587 0.12882650 1.190691 2731
## [2] {ideology=Center,
## life_sat_cat=High,
## redist_opinon=Agree,
## edu_cat=Medium} => {income_feeling=Coping} 0.07337837 0.5293478 0.13862033 1.183956 2922
## [3] {life_sat_cat=High,
## redist_opinon=Agree,
## born_here=Yes,
## edu_cat=Medium,
## imp_rich=Medium} => {income_feeling=Coping} 0.05130459 0.5290005 0.09698400 1.183180 2043
## [4] {life_sat_cat=High,
## redist_opinon=Agree,
## edu_cat=Medium,
## imp_rich=Medium} => {income_feeling=Coping} 0.05439341 0.5285505 0.10291052 1.182173 2166
## [5] {ideology=Center,
## health_cat=Good,
## redist_opinon=Agree,
## born_here=Yes,
## unemployed=No,
## edu_cat=Medium} => {income_feeling=Coping} 0.05012431 0.5252632 0.09542704 1.174821 1996
## [6] {redist_opinon=Agree,
## born_here=Yes,
## unemployed=No,
## edu_cat=Medium,
## imp_rich=Medium} => {income_feeling=Coping} 0.05630195 0.5239542 0.10745586 1.171893 2242
## [7] {life_sat_cat=High,
## gov_sat_cat=Low,
## redist_opinon=Agree,
## born_here=Yes,
## edu_cat=Medium} => {income_feeling=Coping} 0.05047588 0.5235738 0.09640642 1.171042 2010
## [8] {life_sat_cat=High,
## gov_sat_cat=Low,
## redist_opinon=Agree,
## edu_cat=Medium} => {income_feeling=Coping} 0.05286156 0.5232414 0.10102710 1.170298 2105
## [9] {life_sat_cat=High,
## gov_sat_cat=Medium,
## redist_opinon=Agree,
## edu_cat=Medium} => {income_feeling=Coping} 0.05288667 0.5232298 0.10107732 1.170273 2106
## [10] {health_cat=Good,
## redist_opinon=Agree,
## born_here=Yes,
## edu_cat=Medium,
## imp_rich=Medium} => {income_feeling=Coping} 0.05358981 0.5231674 0.10243339 1.170133 2134
Targeted Rules for “Comfortable” Income
The analysis of the ‘Comfortable’ income category identifies a robust segment of success within the European population. With Confidence levels exceeding 70% and Lift values above 2.0, these rules define the protective profile against financial strain.
The dominant predictors of financial comfort are high educational attainment (edu_cat=High), stable employment (unemployed=No), and good subjective health. Interestingly, this group exhibits a clear ideological pattern: high satisfaction with national governance coupled with an opposition to income redistribution (redist_opinon=Disagree). These results suggest that for the most affluent segment, financial security is deeply intertwined with personal health capital, high educational attainment, and a positive outlook on institutional performance.
rules_comfortable <- apriori(data = data_t,
parameter = list(supp = 0.01, conf = 0.6, minlen = 2),
appearance = list(default = "lhs",
rhs = "income_feeling=Comfortable"),
control = list(verbose = F))
inspect(head(sort(rules_comfortable, by = "lift"), 10))## lhs rhs support confidence coverage lift count
## [1] {life_sat_cat=High,
## health_cat=Good,
## redist_opinon=Disagree,
## unemployed=No,
## edu_cat=High,
## imp_helping=High} => {income_feeling=Comfortable} 0.01125035 0.7356322 0.01529344 2.149201 448
## [2] {life_sat_cat=High,
## health_cat=Good,
## redist_opinon=Disagree,
## born_here=Yes,
## unemployed=No,
## edu_cat=High,
## imp_helping=High} => {income_feeling=Comfortable} 0.01009518 0.7322404 0.01378670 2.139292 402
## [3] {life_sat_cat=High,
## gov_sat_cat=High,
## religiosity_cat=Low,
## health_cat=Good,
## unemployed=No,
## edu_cat=High} => {income_feeling=Comfortable} 0.01009518 0.7230216 0.01396248 2.112358 402
## [4] {life_sat_cat=High,
## redist_opinon=Disagree,
## born_here=Yes,
## unemployed=No,
## edu_cat=High,
## imp_helping=High} => {income_feeling=Comfortable} 0.01200372 0.7220544 0.01662439 2.109532 478
## [5] {life_sat_cat=High,
## redist_opinon=Disagree,
## unemployed=No,
## edu_cat=High,
## imp_helping=High} => {income_feeling=Comfortable} 0.01330956 0.7201087 0.01848271 2.103848 530
## [6] {life_sat_cat=High,
## gov_sat_cat=High,
## religiosity_cat=Low,
## born_here=Yes,
## unemployed=No,
## edu_cat=High} => {income_feeling=Comfortable} 0.01009518 0.7140320 0.01413827 2.086094 402
## [7] {life_sat_cat=High,
## health_cat=Good,
## redist_opinon=Disagree,
## unemployed=No,
## edu_cat=High} => {income_feeling=Comfortable} 0.01577057 0.7120181 0.02214912 2.080211 628
## [8] {life_sat_cat=High,
## gov_sat_cat=High,
## religiosity_cat=Low,
## unemployed=No,
## edu_cat=High} => {income_feeling=Comfortable} 0.01172748 0.7118902 0.01647372 2.079837 467
## [9] {health_cat=Good,
## redist_opinon=Disagree,
## unemployed=No,
## edu_cat=High,
## imp_helping=High} => {income_feeling=Comfortable} 0.01197860 0.7098214 0.01687552 2.073793 477
## [10] {gov_sat_cat=High,
## religiosity_cat=Low,
## health_cat=Good,
## unemployed=No,
## edu_cat=High} => {income_feeling=Comfortable} 0.01037141 0.7084048 0.01464052 2.069654 413
Conclusion
The association rules analysis conducted on the ESS Round 11 dataset reveals a clear divide in the socio-economic profiles of European households. By setting income_feeling as the rule consequent, we identified that financial well-being is not a random occurrence but a predictable outcome of specific clusters of life circumstances. The most striking finding is the segmeny associated with the Very Difficult category, where the intersection of labor market exclusion (unemployment) and poor physical health highly increases the likelihood of extreme financial strain. This suggests that for the most vulnerable, economic hardship is deeply intertwined with physical and professional marginalization.
In contrast, the Comfortable and Coping categories represent the more integrated segments of society. While the “Coping” status represents the majority characterized by medium education and stable employment, the “Comfortable” status is strongly protected by high human capital (education) and a high level of institutional satisfaction. Interestingly, an ideological shift is observable across the income feelings: as financial security increases, support for government-led redistribution decreases, and institutional trust significantly rises. Ultimately, these rules demonstrate that subjective poverty in Europe is a multidimensional phenomenon where education, health, and trust in national institutions play just as critical a role as traditional employment status.
| Income_Category | Socio_Economic_Profile | Statistical_Measures | Main_Driver |
|---|---|---|---|
| Very Difficult | Unemployed + Poor Health + Low Trust/Satisfaction | Highest Lift (~9.0); extreme risk cluster | Health & Labor Exclusion |
| Difficult | Low Education + Working + Support for Redistribution | Moderate Lift (~2.5); educational barriers | Low Human Capital |
| Coping | Medium Education + Born in Country + Stable Employment | Baseline Support (~45%); societal norm | Economic Stability |
| Comfortable | High Education + High Trust + Opposition to Redistribution | High Confidence (>70%); protective factors | High Capital & Institutional Trust |