Social segregation in Core Discussion Networks

---

LISS

• We use 14 waves : [2008,2021]
  
• We use 14.580 unique respondents
  
• We use 99.495 respondent-wave combinations
  
• We use 231.020 confidant-respondent-wave combinations

---

Dependent variables

Core Discussion Network

We would like to concentrate on your closest contacts now, to form a picture of the social relationships that people have.
It is easier to answer these types of questions by considering concrete persons.
For that reason, we ask that you list a number of persons close to you. If you wish, you can enter nicknames or initials, as long as you can remember who they refer to.
Most people discuss important things with other people.
If you look back on the last six months, with whom did you discuss important things?
Please enter their first names below (to a maximum of 5).

Figure 4. CDN

CDN homogeneity

Operationalization 1.

To measure the degree of homogeneity within the CDN we use the (reversed) Krackhard and Stern’s E-I index (Krackhardt and Stern 1988). This measure captures the relative prevalence of between-group ties (\(E\)) and within-group ties (\(I\)). It can thus be interpreted as a measure of network segregation.

E-I Index: \(EI = - \frac{E-I}{E+I}\)

We use a reversed version, so higher scores indicate more homogeneity:

• 1 is maximum homogeneity
  
• -1 is maximum heterogeneity

For Figure 4:

• Color homogeneity = \(- \frac{2-3}{2+3}\) = +0.2.
  
• Shape homogeneity = \(- \frac{3-2}{2+3}\) = -0.2.

Operationalization 2.

Similarity Index

S Index = \(S = 2*(\frac{\Sigma_j sim^z_{ij}}{\Sigma_j}) - 1\),
where \(sim^z_{ij} = 1 - \frac{|z_i - z_j|}{r_z}\).

Higer scores indicate more homogeneity/similarity:

• 1 is maximum similarity
  
• -1 is maximum dissimilarity

---

Independent variables

• Education:
  • low (primary, vmbo)
    
  • medium (MBO, HAVO, VWO)
    
  • Professional college (HBO)
    
  • University
    
• Gender:
  • male
    
  • female
    
• Age:
  • birth year
    
  • five year categories
    
• Ethnicity:
  • native Dutch
    
  • non-western
    
  • western

---

Representivity

LISS representative with respect to eductional level measured in five categories?

load(file = "./data/liss_merged_core_file_v3_0921.Rdata")

sample_val <- NA
sample_val[1] <- (table(liss_wide$oplcat.11))[1]
sample_val[2] <- (table(liss_wide$oplcat.11))[2]
sample_val[3] <- (table(liss_wide$oplcat.11))[3] + (table(liss_wide$oplcat.11))[4]
sample_val[4] <- (table(liss_wide$oplcat.11))[5]
sample_val[5] <- (table(liss_wide$oplcat.11))[6]


# cbs https://opendata.cbs.nl/#/CBS/nl/dataset/82816NED/table?ts=1646133108517
datacbs <- c(1265, 2847, 5211, 3122, 1850)
levels(datacbs) <- c("primair", "vmbo", "havo/vwo/mbo", "hbo", "wo")
# round(datacbs / sum(datacbs),2)
pop_val <- datacbs/sum(datacbs)

test <- chisq.test(x = sample_val, p = pop_val, rescale.p = TRUE)
test_data <- data.frame(round(test$observed), round(test$expected), c("primair", "vmbo", "havo/vwo/mbo",
    "hbo", "wo"))
names(test_data) <- c("observed (LISS)", "expected (CBS)", "levels")
test_data
test

rm(list = ls())

#>   observed (LISS) expected (CBS)       levels
#> 1             510            619      primair
#> 2            1424           1393         vmbo
#> 3            2460           2550 havo/vwo/mbo
#> 4            1732           1528          hbo
#> 5             869            905           wo
#> 
#>  Chi-squared test for given probabilities
#> 
#> data:  sample_val
#> X-squared = 51.825, df = 4, p-value = 1.501e-10

Nope, but nothing really problematic.

---

Some data wrangling

load(file="./data/liss_jca_jochem_V2.rds")


liss_long %>%
  mutate(#size
        #cdn_size = 5 - rowSums(is.na(cbind(alter_id_1, alter_id_2, alter_id_3, alter_id_4, alter_id_5))),
        cdn_size = 5 - rowSums(is.na(cbind(gender_alter1, gender_alter2, gender_alter3, gender_alter4, gender_alter5))),
        #education
        opl_y = recode(as.numeric(oplmet), '1' = 6, '2'=10, '3'= 11.5, '4' = 10.5, '5' = 15, '6'=16, '8'=4, .default=-1 ),
        opl_y = na_if(opl_y, -1), 
        opl4 = recode(as.numeric(oplmet), '1' = 1, '2'=1, '3'= 2, '4' = 2, '5' = 3, '6'=4, .default=-1 ),
        opl4 = na_if(opl4, -1), 
        opl4 = structure(opl4, labels=c("low", "medium", "high1", "high2")),
         across(c(educ_recode_alter1,educ_recode_alter2,educ_recode_alter3,educ_recode_alter4,educ_recode_alter5), ~as.numeric(.x), .names = "{.col}_opl4"),
         across(c(educ_recode_alter1_opl4,educ_recode_alter2_opl4,educ_recode_alter3_opl4,educ_recode_alter4_opl4,educ_recode_alter5_opl4), ~recode(.x, '4'=1, '6'=1, '10'=2, '10.5'=2, '11.5'=2 , '15'=3, '16' = 4, .default=-1)),
         across(c(educ_recode_alter1_opl4,educ_recode_alter2_opl4,educ_recode_alter3_opl4,educ_recode_alter4_opl4,educ_recode_alter5_opl4), ~na_if(.x, -1)),
        educ_I = rowSums(cbind(educ_recode_alter1_opl4,educ_recode_alter2_opl4,educ_recode_alter3_opl4,educ_recode_alter4_opl4,educ_recode_alter5_opl4)==opl4,na.rm=T),
         educ_E = rowSums(cbind(educ_recode_alter1_opl4,educ_recode_alter2_opl4,educ_recode_alter3_opl4,educ_recode_alter4_opl4,educ_recode_alter5_opl4)!=opl4,na.rm=T),
         educ_EI =  - ((educ_E - educ_I) / (educ_E + educ_I)),
         cdn_neduc_h = rowSums(cbind(educ_recode_alter1, educ_recode_alter2, educ_recode_alter3, educ_recode_alter4, educ_recode_alter5)==16, na.rm=T),
        cdn_size_educ = 5 - rowSums(is.na(cbind(educ_recode_alter1,educ_recode_alter2,educ_recode_alter3,educ_recode_alter4,educ_recode_alter5))), 
        educ_sim = 2*(1- rowSums(abs((cbind(educ_recode_alter1,educ_recode_alter2,educ_recode_alter3,educ_recode_alter4,educ_recode_alter5) - opl_y)/12), na.rm=T)/cdn_size_educ)-1,
        
        
        #gender   
         gender_alter1 = replace_na(gender_alter1, -1),
         gender_alter2 = replace_na(gender_alter2, -1),
         gender_alter3 = replace_na(gender_alter3, -1),
         gender_alter4 = replace_na(gender_alter4, -1),
         gender_alter5 = replace_na(gender_alter5, -1),
         cdn_ngender_2 = rowSums(cbind(gender_alter1, gender_alter2, gender_alter3, gender_alter4, gender_alter5)==2, na.rm=T),
         cdn_ngender_1 = rowSums(cbind(gender_alter1, gender_alter2, gender_alter3, gender_alter4, gender_alter5)==1, na.rm=T),
         gender_I = ifelse(geslacht==2, cdn_ngender_2, cdn_ngender_1),
         gender_E = ifelse(geslacht==1, cdn_ngender_2, cdn_ngender_1),
         gender_EI =  - ((gender_E - gender_I) / (gender_E + gender_I)),
        
        #age
        across(c(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5), ~na_if(.x, 14)),
        across(c(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5), ~as.integer(.x)),
        leeftijd_cat13 = as.integer(cut(as.numeric(liss_long$leeftijd),breaks = c(-Inf, 15, 20, 25, 30, 35, 40,45,50,55,60,65,70, Inf))),
        leeftijd_I = rowSums(cbind(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5)==leeftijd_cat13, na.rm=T),
        leeftijd_E = rowSums(cbind(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5)!=leeftijd_cat13, na.rm=T),
        leeftijd_EI =  - ((leeftijd_E - leeftijd_I) / (leeftijd_E + leeftijd_I)),
        
        cdn_size_age = 5 - rowSums(is.na(cbind(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5))), 
        across(c(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5), ~recode(.x, '1'=15, '2'=18, '3'=23, '4'=28, '5'=33 , '6'=38, '7' = 43,'8'= 48, '9'=53, '10'=58, '11'=63, '12'=68, '13'=75, .default=-1)),
         across(c(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5), ~na_if(.x, -1)),
        leeftijd_sim = 2*(1 - rowSums(abs((cbind(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5) - as.numeric(liss_long$leeftijd))/75), na.rm=T)/cdn_size_age)-1,
        
        
        
        #etni
        etni3 = recode(as.numeric(origin), '0' = 1, '101'=2, '102'= 3, '201' = 2, '202' = 3, '999'=-1, .default=-1 ),
        etni3 = na_if(etni3, -1), 
        etni3 = structure(etni3, labels=c("D", "W", "NW")),
         across(c(origin_alter1,origin_alter2,origin_alter3,origin_alter4,origin_alter5), ~as.numeric(.x), .names = "{.col}_etni3"),
         across(c(origin_alter1_etni3,origin_alter2_etni3,origin_alter3_etni3,origin_alter4_etni3,origin_alter5_etni3), ~recode(.x, '1'=1, '2'=3, '3'=3, '4'=3, '5'=3 , '6'=2, '7' = 3, '8' = 2, '9' = -1, .default=-1)),
         across(c(origin_alter1_etni3,origin_alter2_etni3,origin_alter3_etni3,origin_alter4_etni3,origin_alter5_etni3), ~na_if(.x, -1)),
         etni_I = rowSums(cbind(origin_alter1_etni3, origin_alter2_etni3, origin_alter3_etni3, origin_alter4_etni3 ,origin_alter5_etni3)==etni3, na.rm=T),
         etni_E = rowSums(cbind(origin_alter1_etni3,origin_alter2_etni3,origin_alter3_etni3,origin_alter4_etni3,origin_alter5_etni3)!=etni3, na.rm=T),
         etni_EI =  - ((etni_E - etni_I) / (etni_E + etni_I)),
        
          ego_id = nomem_encr) %>%
  #filter(!is.na(pol_int))   %>%
  select(ego_id,  survey_wave, leeftijd, leeftijd_cat13,  opl4, geslacht, etni3, cdn_size, cdn_size_educ, cdn_size_age, educ_EI,educ_sim,gender_EI, leeftijd_EI, leeftijd_sim, etni_EI ) -> datajt

#make similar dataset for non-kin. dus gooi alle alters weg met rel_alter - 1 2,3,4,5
# quickest would probably be to go from wide to long, select and go back to wide. but...

attributes(liss_wide$rel_alter1.1)

liss_long %>%
  mutate(across(c(rel_alter1,rel_alter2,rel_alter3,rel_alter4,rel_alter5), ~replace_na(.x, -1)),
         #alter_id_1 = ifelse(rel_alter1>0 & rel_alter1<6, NA, alter_id_1), 
         #alter_id_2 = ifelse(rel_alter2>0 & rel_alter2<6, NA, alter_id_2), 
         #alter_id_3 = ifelse(rel_alter3>0 & rel_alter3<6, NA, alter_id_3), 
         #alter_id_4 = ifelse(rel_alter4>0 & rel_alter4<6, NA, alter_id_4), 
         #alter_id_5 = ifelse(rel_alter5>0 & rel_alter5<6, NA, alter_id_5), 
         
         educ_recode_alter1 = ifelse(rel_alter1>0 & rel_alter1<6, NA, educ_recode_alter1), 
         educ_recode_alter2 = ifelse(rel_alter2>0 & rel_alter2<6, NA, educ_recode_alter2), 
         educ_recode_alter3 = ifelse(rel_alter3>0 & rel_alter3<6, NA, educ_recode_alter3), 
         educ_recode_alter4 = ifelse(rel_alter4>0 & rel_alter4<6, NA, educ_recode_alter4), 
         educ_recode_alter5 = ifelse(rel_alter5>0 & rel_alter5<6, NA, educ_recode_alter5), 
         
         gender_alter1 = ifelse(rel_alter1>0 & rel_alter1<6, NA, gender_alter1), 
         gender_alter2 = ifelse(rel_alter2>0 & rel_alter2<6, NA, gender_alter2), 
         gender_alter3 = ifelse(rel_alter3>0 & rel_alter3<6, NA, gender_alter3), 
         gender_alter4 = ifelse(rel_alter4>0 & rel_alter4<6, NA, gender_alter4), 
         gender_alter5 = ifelse(rel_alter5>0 & rel_alter5<6, NA, gender_alter5),
         
         origin_alter1 = ifelse(rel_alter1>0 & rel_alter1<6, NA, origin_alter1), 
         origin_alter2 = ifelse(rel_alter2>0 & rel_alter2<6, NA, origin_alter2), 
         origin_alter3 = ifelse(rel_alter3>0 & rel_alter3<6, NA, origin_alter3), 
         origin_alter4 = ifelse(rel_alter4>0 & rel_alter4<6, NA, origin_alter4), 
         origin_alter5 = ifelse(rel_alter5>0 & rel_alter5<6, NA, origin_alter5),
         
         age_alter1 = ifelse(rel_alter1>0 & rel_alter1<6, NA, age_alter1), 
         age_alter2 = ifelse(rel_alter2>0 & rel_alter2<6, NA, age_alter2), 
         age_alter3 = ifelse(rel_alter3>0 & rel_alter3<6, NA, age_alter3), 
         age_alter4 = ifelse(rel_alter4>0 & rel_alter4<6, NA, age_alter4), 
         age_alter5 = ifelse(rel_alter5>0 & rel_alter5<6, NA, age_alter5),
         
         
         ) %>%


  mutate(#size
        #cdn_size = 5 - rowSums(is.na(cbind(alter_id_1, alter_id_2, alter_id_3, alter_id_4, alter_id_5))),
        cdn_size = 5 - rowSums(is.na(cbind(gender_alter1, gender_alter2, gender_alter3, gender_alter4, gender_alter5))),
        #education
        opl_y = recode(as.numeric(oplmet), '1' = 6, '2'=10, '3'= 11.5, '4' = 10.5, '5' = 15, '6'=16, '8'=4, .default=-1 ),
        opl_y = na_if(opl_y, -1), 
        opl4 = recode(as.numeric(oplmet), '1' = 1, '2'=1, '3'= 2, '4' = 2, '5' = 3, '6'=4, .default=-1 ),
        opl4 = na_if(opl4, -1), 
        opl4 = structure(opl4, labels=c("low", "medium", "high1", "high2")),
         across(c(educ_recode_alter1,educ_recode_alter2,educ_recode_alter3,educ_recode_alter4,educ_recode_alter5), ~as.numeric(.x), .names = "{.col}_opl4"),
         across(c(educ_recode_alter1_opl4,educ_recode_alter2_opl4,educ_recode_alter3_opl4,educ_recode_alter4_opl4,educ_recode_alter5_opl4), ~recode(.x, '4'=1, '6'=1, '10'=2, '10.5'=2, '11.5'=2 , '15'=3, '16' = 4, .default=-1)),
         across(c(educ_recode_alter1_opl4,educ_recode_alter2_opl4,educ_recode_alter3_opl4,educ_recode_alter4_opl4,educ_recode_alter5_opl4), ~na_if(.x, -1)),
        educ_I = rowSums(cbind(educ_recode_alter1_opl4,educ_recode_alter2_opl4,educ_recode_alter3_opl4,educ_recode_alter4_opl4,educ_recode_alter5_opl4)==opl4,na.rm=T),
         educ_E = rowSums(cbind(educ_recode_alter1_opl4,educ_recode_alter2_opl4,educ_recode_alter3_opl4,educ_recode_alter4_opl4,educ_recode_alter5_opl4)!=opl4,na.rm=T),
         educ_EI =  - ((educ_E - educ_I) / (educ_E + educ_I)),
         cdn_neduc_h = rowSums(cbind(educ_recode_alter1, educ_recode_alter2, educ_recode_alter3, educ_recode_alter4, educ_recode_alter5)==16, na.rm=T),
        cdn_size_educ = 5 - rowSums(is.na(cbind(educ_recode_alter1,educ_recode_alter2,educ_recode_alter3,educ_recode_alter4,educ_recode_alter5))), 
        educ_sim = 2*(1- rowSums(abs((cbind(educ_recode_alter1,educ_recode_alter2,educ_recode_alter3,educ_recode_alter4,educ_recode_alter5) - opl_y)/12), na.rm=T)/cdn_size_educ)-1,
        
        
        #gender   
         gender_alter1 = replace_na(gender_alter1, -1),
         gender_alter2 = replace_na(gender_alter2, -1),
         gender_alter3 = replace_na(gender_alter3, -1),
         gender_alter4 = replace_na(gender_alter4, -1),
         gender_alter5 = replace_na(gender_alter5, -1),
         cdn_ngender_2 = rowSums(cbind(gender_alter1, gender_alter2, gender_alter3, gender_alter4, gender_alter5)==2, na.rm=T),
         cdn_ngender_1 = rowSums(cbind(gender_alter1, gender_alter2, gender_alter3, gender_alter4, gender_alter5)==1, na.rm=T),
         gender_I = ifelse(geslacht==2, cdn_ngender_2, cdn_ngender_1),
         gender_E = ifelse(geslacht==1, cdn_ngender_2, cdn_ngender_1),
         gender_EI =  - ((gender_E - gender_I) / (gender_E + gender_I)),
        
        #age
        across(c(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5), ~na_if(.x, 14)),
        across(c(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5), ~as.integer(.x)),
        leeftijd_cat13 = as.integer(cut(as.numeric(liss_long$leeftijd),breaks = c(-Inf, 15, 20, 25, 30, 35, 40,45,50,55,60,65,70, Inf))),
        leeftijd_I = rowSums(cbind(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5)==leeftijd_cat13, na.rm=T),
        leeftijd_E = rowSums(cbind(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5)!=leeftijd_cat13, na.rm=T),
        leeftijd_EI =  - ((leeftijd_E - leeftijd_I) / (leeftijd_E + leeftijd_I)),
        
        cdn_size_age = 5 - rowSums(is.na(cbind(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5))), 
        across(c(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5), ~recode(.x, '1'=15, '2'=18, '3'=23, '4'=28, '5'=33 , '6'=38, '7' = 43,'8'= 48, '9'=53, '10'=58, '11'=63, '12'=68, '13'=75, .default=-1)),
         across(c(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5), ~na_if(.x, -1)),
        leeftijd_sim = 2*( 1 - rowSums(abs((cbind(age_alter1, age_alter2, age_alter3, age_alter4, age_alter5) - as.numeric(liss_long$leeftijd))/75), na.rm=T)/cdn_size_age)-1,
        
        
        
        #etni
        etni3 = recode(as.numeric(origin), '0' = 1, '101'=2, '102'= 3, '201' = 2, '202' = 3, '999'=-1, .default=-1 ),
        etni3 = na_if(etni3, -1), 
        etni3 = structure(etni3, labels=c("D", "W", "NW")),
         across(c(origin_alter1,origin_alter2,origin_alter3,origin_alter4,origin_alter5), ~as.numeric(.x), .names = "{.col}_etni3"),
         across(c(origin_alter1_etni3,origin_alter2_etni3,origin_alter3_etni3,origin_alter4_etni3,origin_alter5_etni3), ~recode(.x, '1'=1, '2'=3, '3'=3, '4'=3, '5'=3 , '6'=2, '7' = 3, '8' = 2, '9' = -1, .default=-1)),
         across(c(origin_alter1_etni3,origin_alter2_etni3,origin_alter3_etni3,origin_alter4_etni3,origin_alter5_etni3), ~na_if(.x, -1)),
         etni_I = rowSums(cbind(origin_alter1_etni3, origin_alter2_etni3, origin_alter3_etni3, origin_alter4_etni3 ,origin_alter5_etni3)==etni3, na.rm=T),
         etni_E = rowSums(cbind(origin_alter1_etni3,origin_alter2_etni3,origin_alter3_etni3,origin_alter4_etni3,origin_alter5_etni3)!=etni3, na.rm=T),
         etni_EI =  - ((etni_E - etni_I) / (etni_E + etni_I)),
        
          ego_id = nomem_encr) %>%
  #filter(!is.na(pol_int))   %>%
  select(ego_id,  survey_wave, leeftijd, leeftijd_cat13,  opl4, geslacht, etni3, cdn_size, cdn_size_educ, cdn_size_age, educ_EI,educ_sim,gender_EI, leeftijd_EI, leeftijd_sim, etni_EI )  -> datajt_nk




test <- cut(as.numeric(liss_long$leeftijd),breaks = c(-Inf, 15, 20, 25, 30, 35, 40,45,50,55,60,65,70, Inf))
#levels(test)

save(datajt, file="./data/datajt.rda")
save(datajt_nk, file="./data/datajt_nk.rda")

#cor.test(datajt$educ_EI, datajt$educ_sim)
#cor.test(datajt$leeftijd_EI, datajt$leeftijd_sim)

#hist(datajt$leeftijd_sim)

#rm(list=ls())  

---

References

Krackhardt, David, and Robert N Stern. 1988. “Informal Networks and Organizational Crises: An Experimental Simulation.” Social Psychology Quarterly, 123–40.

---

Copyright © 2022 Jochem Tolsma