Perceptions of Class Conflicts

In this blog post I will investigate peoples perception of class conflicts in Sweden between 1992 and 2009. The data I am going to use is International Social Survey Project (ISSP). The data is free to download — but you need to register. Data can be found here: https://www.gesis.org/en/issp/home.

To investigate class conflicts is interesting and the context is an assumed decrease in understanding class relations as conflicts (see for example Pakulski & Waters, 1996) in the population, but also since class has become popular once again (in Sweden and in Swedish see for example Suhonen et al., 2021, in Britain see for example Evans & Tilley, 2017) at least among academics and intellectuals. But class is mainly of interest since class is understood to be of importance for the welfare state and the structure of the society in general (see for example Esping-Andersen, 1990).

Here I will investigate four different kinds of class conflict: conflict between workers and managers, conflict between poor and rich, conflict between the top and the bottom and conflict between working class and the middle class. The hypothesis is that class conflict is understood as less concrete and more abstract (less and less understands conflicts as based on workers and managers and working class and middle class, and more and more understands conflicts as between the poor and the rich and top and bottom).

To investigate this I will use R statistics. Since the data is downloaded as a .sav-file (used by SPSS) I need a tool that can read it into R. There are several, but here I will use haven. Haven has a really simple code for reading SPSS-files into R:

library (haven)
wd <- read_sav("ZA2310_2006-07-18.sav")

The data sets that will be used is from 1992, 1998 and 2009. The reason for these years are that ISSP had inequality as a theme these year and the class conflict questions is only available in this theme. The strategy is to load each year into to different datasets:

d1992 <- read_sav("ZA2310_2006-07-18.sav")
d1999 <- read_sav("ZA3430_2006-09-14.sav")
d2009 <- read_sav("ZA5400_v4-0-0.sav")

The subset-commando are used to separate Sweden from the other countries (i.e. only using Sweden):

s1992 <- subset(d1992, V3==10)
s1999 <- subset(d1999, V3==13)
s2009 <- subset(d2009, V4==752)

It is further practical to use lowercase/common letters.

names(s1992) <- tolower (names(s1992))
names(s1999) <- tolower (names(s1999))
names(s2009) <- tolower (names(s2009))

There are several ways and methods to transform and manipulate variables in R. Most popular is probably to use //. I have actually never got into using Tidyverse, even though I use some of the packages included (as haven above and ggplot — which I think is part of Tidyverse). I have nothing against Tidyverse, just have not taken time to get into it. But I will probably write a blog post about this in the future. Here I will use the car package and the recode function. To load car just:

library(car)

Of course you need to install the package first.

install.packages("car")

I will start by recoding the worker and manager conflict. The conflict variables all look the same and have four categories:

1. Very strong conflict
2. Strong conflicts
3. Not very strong conflicts
4. No conflicts

One issue with the recode function in car is that it can mess up the order of the variables. The reason is that it orders the variables alphabetical or in accordance to what number comes first. To deal with this we can use the level command inside the recode function

s1992$worker_managers <- recode(s1992$v68,'
1 = "Very strong conflicts";
2 = "Strong conflicts";
3 = "Not very strong conflicts";
4 = "No conflicts"',
as.factor=TRUE,
levels=c("Very strong conflicts",
         "Strong conflicts",
         "Not very strong conflicts",
         "No conflicts"))

However, it is complicated to use so many categories in the analysis. The results will be easier to understand if the variable is simplified. Here I have dichotomized the variable.

s1992$wm <- recode(s1992$v68,'
1:2 = "Conflict";
3:4 = "No conflict"')

The other three variables have been transformed in the same way.

No we can look at some results. In this case it is rather easy, just doing some simple tables — for example:

t <- table(s1999$worker_managers)
cbind (round(prop.table(t)*100, 1))
t

The result is presented in Table 1.

It seems as perceptions of class conflict based on this one was substantially lower in 1992 than in 1999. The differences between 1999 and 2009 are smaller, but it seems as fewer sees the relation between workers and managers as a conflict relation in 2009. However, the surprising result here is the difference between 1992 and 1999. This can of course be a result of some kind of problem with the data. But maybe perception of conflict were lower in the beginning of the 1990’s. Sweden had a relatively new centre-right wing government lead by the right wing party Moderaterna. There was also an economic crisis going on — which maybe in the beginning resulted in some kind of crisis consciousness. I don’t know, but this makes analyses on the other types of conflict even more interesting.

In table 2 perceptions of conflict between rich and poor are presented. As can be seen the general pattern is similar in the sence that the proportion understanding that it is a conflict between poor and rich are smallest in 1992. However, the proportion does not decrease between 1999 and 2009, and the differences between the years are rather small.

Next conflict type to investigate is between top and bottom. unfortunately, this variable does not exist for year 1992. Anyway, the results for 1999 and 2009 are presented in Table 3.

From Table 3 it seems as the proportion that thinks there is a conflict between the bottom and the top in society increased somewhat between 1999 and 2009. However, the difference between the years is small. So when investigating two more abstract forms of conflict the pattern is somewhat different, and the volumen is substantially different. How about another more concrete conflict type — here conflict between wokring class and middle class.

Well, this result is interesting. The results here resembles the results in Table 1. But when looking closer there is no increase in perceptions of strong conflicts between 1992 and 1999, which was the case regarding conflict between workers and managers. In fact there is a decrease regarding both very strong conflict and strong conflict. Fewer means that there are no conflicts, while more means that there are conflicts, but not very strong. More countries should be investigated, but it is clear that different kind of class conflicts develops differently, and it seems as class conflict based on concrete conflicts has a different pattern than class conflict based on more abstract conflict pattern. Further, it seems as perception of concrete class conflict increased during the 1990’s.

To get at better view I will put the result into a graph. To do that I have dichotomized the variables, by separating very strong conflicts and strong conflicts from not very strong conflict and no conflict.