Deputies preparation

We define the function for preparing the deputies:

prep_deputies <- function(deputies, end_date = "2021-12-02") {

deputies <- deputies %>%

# removing office term included in the party name var party is provide as
# "PartyX (1900-00-01-[...])" we need to remove what's after the first "("
separate(col = partito,
         into = c("party", "trash"),
         sep = "\\(") %>%
mutate(
  # arty column has space after the party name, then we must trim it!
  party = str_trim(party),

  # parsing offices data into a single interval format
  date = interval(start = ymd(s_office),
                  end = ymd(ifelse(test = is.na(deputies$e_office),
                                   #since the 18th leg is still in office
                                   #some MPs does not have ending date
                                   yes = 20220228,
                                   no = deputies$e_office)))) %>%

# dropping working variables no more useful
select(!c(trash, s_office, e_office)) %>%

# drop deputies which have taken office *AFTER* Conte_ii cabinet
filter(int_start(date) < lubridate::ymd(end_date))

#reuturn value
return(deputies)
}

The function’s primary purpose is to deal with the query’s problematic results. The query returned four raw columns:

• name of the deputy
• party of the deputy
• office start date
• office end date

The major problem is the party column because it includes the party name and the terms of the mandate, written between round brackets. With separate(), we can split the party name from the terms, which are unused since we have it already. Then, we parse the start and end dates into an interval object.

Contributions preparation

prep_cabinet <- function(cabinet, cabinet_int, deputies) {
`%!in%` <- Negate(`%in%`) # "not in" function declaration

#define cabinet name
name <- as.character(substitute(cabinet))  


# Due to the construction of dati.camera.it database it's easier preparing
#data within R environment:
cabinet <- cabinet %>%
 
 # removing duplicate rows
 distinct(.) %>%
 
 # Parsing dates
 mutate(date = ymd(date))

# Extracting deputies and contributors
cabinet_deputies <- deputies %>%
 filter(int_start(date) < int_end(cabinet_int)) %>%
 filter(!(int_end(date) < int_start(cabinet_int)))

# Since some MPs, may have changed party or decayed in between
# they must be coded as SWITCHER/DECAYED
is_duplicate <- cabinet_deputies %>%
 
 # group by MPs names
 group_by(name) %>%
 
 # n as how many time a unique name appears
 summarise(n = n()) %>%
 
 # filter only the names who appear more than one time
 filter(n > 1) %>%
 
 # coding party as "SWITCHER/DECAYED"
 add_column(party = "SWITCHER/DECAYED") %>%
 
 #selecting only name an party to match the structure of cabinet_deputies
 select(name, party)


cabinet_deputies <- cabinet_deputies %>%
 
 # remove switcher from cabinet_deputies with an anti joint fun
 anti_join(is_duplicate, by = "name") %>%
 
 #drop the date column, no more useful
 select(!(date)) %>%
 
 #bind the Switcher to cabinet_deputies
 bind_rows(., is_duplicate)

#remove eventual orphan nodes
cabinet_deputies <-  cabinet_deputies %>% 
  filter(!(name %!in% c(cabinet$signatory,cabinet$joint_signatory)))

# since there is MPs no included in the nodes df we must debug it!
debug <- cabinet[which(!(cabinet$joint_signatory %in% cabinet_deputies$name)), ]
debug <- unique(debug$joint_signatory)
cabinet_deputies <- rbind(cabinet_deputies,
                         deputies[deputies$name %in% debug, 1:2])

#assign the contributions DFs
assign(name, 
      cabinet,
      envir = parent.frame())

#assign the deputies DFs
assign(stringr::str_c(name, "_deputies"),
      cabinet_deputies,
      envir = parent.frame())
}

The function consists of two primary purposes: first, removing the duplicates observation on the contributions dataframe. Furthermore, it deals with a significant issue: Italian members of Parliament can change the party during their mandate. This, cause the presence of duplicated names (vertices in the SNA) which must be removed.

The issue has been addressed by filtering for the deputies whose office mandate started before the cabinet end date and dropping the deputies whose office mandate started after the cabinet end date. Then, the deputies who appear more than once are saved in a temporary object, coding the party variables as “SWITCHER/DECAYED”. Lastly, the duplicates deputies are removed from the cabinet_deputies dataframe using an anti_join(), and then they are appended again to the dataframe. This method permits to have solely unique observations in the cabinet_deputies dataframe. Hence, the observations that appeared more than once are now unique, with the party coded as “SWITCHER/DECAYED”

To remove eventual orphan nodes, the function removes the observations from cabinet_deputies dataframe, whose name is not present in the contributions dataframe, cabinet. In order to doing so, we used a user-defined operator: %!in%². The operator returns the values of \(x\) that do not match in \(y\). Applied to our case using a negate filter, it filters out all the deputies who are not present in the contributors’ dataframe³.

The function addresses another issue, also; some bills have latecomers associated as joint signatories since the lengthy legislative procedure. Possibly, these latecomers have not taken office under the cabinet in analysis and must be removed. The solution is implemented in a slightly convoluted manner which must be addressed in future releases.

At its last, the function returns the resulting dataframes to the parent environment using the assign() verb The name of the returned objects is built upon the dataframe given as the cabinet argument to the function.

Defined all the functions, we can now use them finishing the data pre-processing:

# Deputies function
deputies <- prep_deputies(deputies)

# Conte_i contributions and dates
conte_i_date <- interval(ydm("2018-31-05"), ydm("2019-04-09"))

prep_cabinet(cabinet = conte_i, cabinet_int = conte_i_date, deputies = deputies)

# conte_ii contributions and dates
conte_ii_date <- interval(ydm("2019-04-09"), ydm("2021-12-02"))

prep_cabinet(cabinet = conte_ii, cabinet_int = conte_ii_date, deputies = deputies)

Italia Viva

A significant shortcoming was introduced with Conte’s 2nd cabinet. Italia Viva party was founded (2019-09-18) fourteen days after the cabinet was sworn. For this reason, the function drop IV deputies as switchers. For this reason, we must add the deputies manually into the cabinet_deputies dataframe:

iv <- deputies %>%
filter(party == "ITALIA VIVA") %>%
filter(int_start(date) == ymd("2019-09-19") &
    int_end(date) > int_end(conte_ii_date)) %>%
select(!(date))

conte_ii_deputies$party <- ifelse(conte_ii_deputies$name %in% iv$name,
                            "ITALIA VIVA",
                            conte_ii_deputies$party)

Visualization

To visualize the resulting network, we plot it tuning the arguments to obtain the most precise result. We applied an algorithm that spread the nodes using a physical simulation to better position them. For this analysis, we emplyed the Fruchterman-Reingold (1991) algorithm.

# Fruchterman-Reingold layout computation:
l <- layout_with_fr(conte_i_network, niter = 2000)

# Plot network
plot(
  
  ## Network main arguments
  conte_i_network, 
  frame = F,
  layout = l,
  
  ## Vertexes aes:
  vertex.size = 7, 
  vertex.label = NA,
  vertex.label.cex = .7,
  vertex.color = adjustcolor(V(conte_i_network)$color, alpha.f = 0.9),
  
  ## Edges aes: 
  edge.arrow.size = .05,
  edge.width = .3,
  edge.curved =0.3)

CPS

To calculate the Cosponsorship index, we defined a simple function. First, we double joint the deputies df into contributions df to extract the deputies party. Then, grouping by the party of the first signatory, we divided the numbers of joint signatories of the same party by the total of joint signatories that had contributed to the party’s bills:

cosponsor <- function(contributions, 
                      deputies, signatory = "signatory", 
                      joint_signatory = "joint_signatory",
                      mp_name = "name" ) {
   contributions %>% 
      
      # double join to get the signatory and joint signatory party 
      left_join(deputies,
                by = c(signatory = mp_name )) %>% 
      left_join(deputies,
                by = c(joint_signatory = mp_name),
                suffix = c("_main","_joint")) %>% 
      
      # group by the bill and the party of the signatory and computing the csp
      group_by(party_main) %>% 
      summarise(n= n(),
                n_joint = sum(party_joint == party_main)) %>% 
      mutate(csp = (n_joint/n)) %>% 
      summarise(party = party_main,
                csp = csp)
}

IPS

We used the same structure as the previous function to calculate the intraparty index. This time, we grouped by the party of the joint signatories. So, we calculated all bills co-sponsored by MPs of a party, the percentage of those initiated by the same party:

intraparty <- function(contributions, 
                       deputies, signatory = "signatory", 
                       joint_signatory = "joint_signatory",
                       mp_name = "name" ) {
   contributions %>% 
      left_join(deputies,
                by = c(signatory = mp_name )) %>% 
      left_join(deputies,
                by = c(joint_signatory = mp_name),
                suffix = c("_main","_joint")) %>% 
      
      # group by party of the joint signatory
      group_by(party_joint) %>% 
      summarise(n= n(),
                n_joint = sum(party_joint == party_main)) %>% 
      
      # computing the ISP index  
      mutate(ips = (n_joint/n)) %>% 
      summarise(party = party_joint,
                ips = ips)
}

IOBD index

At last, we defined the last function, which calculates the sum of the two indexes defined before, relying on the above-explained functions:

iobd_index <- function(contributions, deputies, 
                 signatory = "signatory", joint_signatory = "joint_signatory",
                 mp_name = "name" ) {
   
   # call cosponsors function
   cosponsor(contributions, deputies, 
             signatory, joint_signatory,
             mp_name) %>%
      
   #join the result with intraparty function by party name
      left_join(., intraparty(contributions, deputies, 
                              signatory, joint_signatory,
                              mp_name),
                by = c("party" = "party")) %>% 
      
   # summarize the results by the party name and IOBD var,
   # which is the sum of CSP and IPS
      summarise(party,
                IOBD = csp+ips)
}

Network analysis

Conte I

The visual analysis of the produced SNA graph permits some considerations. As can be observed in Figure 1, during the first Conte’s cabinet the majority of the parties were sheltered to their positions. Even if the two government parties, Lega Nord and Movimento 5 Stelle, have interacted extensively between them, they still constitute two distinct communities. Only the Partito Democratico results partially blended with Liberi e Uguali community. It must be noted that this visualization did not employ any modularity computation, and the nodes (which represent deputies) has been coloured by the party affiliation.

IOBD index

For the last part of the project, we calculated the IODB index for the two periods under analysis using the defined functions. The results, shown in Table 1, provide evidence in favour of our previous findings.
Under Conte I, all the parties demonstrated strong ties within them. Under Conte II, the parties had strong ties within them; however, the ties between the ruling parties were stronger than Conte I cabinet.