All pages

Chapter 3: Interpreting data clearly

For those curious about levels of well-being of people across Europe, this is where things start getting interesting. Which countries have the highest levels of personal and social well-being? How are the patterns different for different aspects of well-being? In which aspects does a particular country do well or badly? What about differences with respect to different age groups, genders and income levels? What about how distributions vary from country to country? Which countries have high well-being inequality and which have low well-being inequality? The number of questions one can ask is endless and we encourage you to explore them freely. For now we present a few examples of the kinds of things you can do.

Page 1

The big picture

Before breaking down well-being into the various components and sub-components it will help to look at the overall picture. In which countries are personal and social well-being high, in which are they low?

Exercise 7

Simply calculate the means for personal and social well-being by country, using appropriate weighting. If you are using Nesstar, use the map function to visualise this. Which regions of Europe have highest personal and which highest social? Are they differences between the two types? (Note: Calculate these means using untransformed scores. If you wish, you can transform them afterwards but it’s not necessary and won’t change how the countries fare against one another).

Solution - Nesstar
  1. Open the dataset.
  2. Click the icon for weighting, and select the design weight.
  3. Select the tab ‘Tabulation’, find the variable ‘map - Map identification code for countries’ in the variable group ‘Country’ and select ‘Add to row’.
  4. Find the variable ‘Personal well-being’ in the variable group ‘Well-being, recoded and computed variables’, and select ‘Add as measure’.
  5. Sort the table by clicking the column header ‘average’.
  6. Visualise the results by clicking the icon for map.
  7. Repeat 4 – 6 for the variable ‘Social well-being’.

Table 3.1. The rank orders (transformed or otherwise)
Personal Social
Denmark Denmark
Switzerland Norway
Austria Spain
Norway Switzerland
Finland Sweden
Ireland Ireland
Sweden Portugal
Netherlands Netherlands
Cyprus Finland
Belgium Austria
Germany Hungary
Spain Cyprus
UK Belgium
Slovenia Bulgaria
France UK
Poland Estonia
Estonia Germany
Portugal Poland
Slovakia France
Hungary Slovenia
Bulgaria Slovakia
Ukraine Ukraine

As you can see the regions with highest personal well-being are in Scandinavia, as well as Austria and Switzerland, the lowest being towards the East of Europe and Portugal. For social well-being, Scandinavia still does well, but so does Southern Europe. Meanwhile, Central and Eastern Europe, with some countries still doing badly (e.g. Ukraine) whilst others do better (e.g. Hungary).

 

Solution SPSS

Table 3.1. The rank orders (transformed or otherwise)
Personal Social
Denmark Denmark
Switzerland Norway
Austria Spain
Norway Switzerland
Finland Sweden
Ireland Ireland
Sweden Portugal
Netherlands Netherlands
Cyprus Finland
Belgium Austria
Germany Hungary
Spain Cyprus
UK Belgium
Slovenia Bulgaria
France UK
Poland Estonia
Estonia Germany
Portugal Poland
Slovakia France
Hungary Slovenia
Bulgaria Slovakia
Ukraine Ukraine

 

As you can see the regions with highest personal well-being are in Scandinavia, as well as Austria and Switzerland, the lowest being towards the East of Europe and Portugal. For social well-being, Scandinavia still does well, but so does Southern Europe. Meanwhile, Central and Eastern Europe, with some countries still doing badly (e.g. Ukraine) whilst others do better (e.g. Hungary).

WEIGHT BY dweight.
 
MEANS
TABLES=personal_WBI social_WBI BY CNTRY
/CELLS MEAN COUNT STDDEV.
Page 2

Country profiles

One of the most useful approaches to the data is to take a snapshot of a single country and look at how it fares in terms of different aspects of well-being – its well-being profile.

Let us for example take a closer look at Portugal (we have used SPSS):

  1. Subset the data to the Portuguese respondents.
  2. Switch on the design weight.
  3. Calculate the mean scores for the following components: positive feelings, absence of negative feelings, satisfying life, vitality, resilience & self-esteem, positive functioning, supportive relationships, and trust & belonging. If you want to have a more in-depth profile, then you could select further sub-components (for example different aspects of positive functioning).
  4. Copy the table from the output with all the means and paste it into an Excel sheet.
  5. Use Excel to create a ‘radar’ graph, below. From the menu, select ‘Insert’ – ‘Diagram’ - ‘Radar’.

Remember that 0 is the European average for any given component or sub-component.

Exercise 8

Create a ‘radar’ graph similar to the one in the figure above using data from a different country than Portugal.

You can do this with other sub-categories (e.g. age groups). It is also possible to plot more than one profile onto the same graph to allow direct comparisons. Why not calculate mean component scores for males and females for a given country, and plot the results?

SPSS syntax

*SPSS syntax for the example.

WEIGHT BY DWEIGHT.
COMPUTE filter_$=(CNTRY='PT').
VARIABLE LABEL filter_$ "CNTRY='PT' (FILTER)".
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE.
 
DESCRIPTIVES
VARIABLES=positive_affect negative_affect satisfaction vitality selfesteem functionings supportive_relations, trust_belonging
/STATISTICS=MEAN.
 
FILTER OFF.
USE ALL.
EXECUTE .
Page 3

Rolling hills

Another way to clearly visualise differences in components is in a ‘rolling hills’ graph, as shown below. In Figure 6 we present a given component (in this case ‘negative feelings’) and see how the order of countries differs for this component compared to its parent index – personal well-being index. The countries along the bottom are ordered in terms of the personal well-being index (with Denmark first, etc.). That way, it’s easy to spot, not only those countries with good scores on this component (e.g. Finland, Norway and Denmark), but also those countries that do particularly well or badly on this component relative to their overall position. For example, whilst Austria doesn’t do particularly badly (8th highest score), it is easy to see that it does very badly given its overall personal well-being score. In other words, Austrians suffer quite a high level of negative feelings given their overall high personal well-being.

Exercise 9

Try and replicate the graph above with the same or another component, sub-component or even with the z-score for an individual question. If the indicator you have chosen is a subset of personal well-being, then order your countries as we have done, according to their mean personal well-being index. If, however, it is a subset of social well-being, then order them according to their mean social well-being indices.

Solution and SPSS syntax

Once you’ve calculated the means, transfer the scores and country names into Excel. Sort them by the overall scores, such that the country with highest overall index is top of the list (Denmark in both cases), and that with the lowest overall index is bottom (Ukraine). Then plot a graph with just the indicator, not the overall index.

 

WEIGHT BY DWEIGHT.
 
MEANS
TABLES= negative_affect personal_WBI BY CNTRY
/CELLS MEAN.

Exercise 10: Income groups

Let’s do an analysis where the income group, rather than country, is the main independent variable. Take a single country. In the data set, there are two income variables. One retains the original 12 income bands in the ESS data set. The other compresses them into 4 bands so as to increase frequencies per band. We’ll use the latter of these, as there may not be enough data points across all 12 bands in a single country. Calculate mean values for a few components across the different income bands. If plotting these results, use the confidence intervals option to help draw conclusions (this is particularly important if you do use all 12 bands because it will help identify where there are few data points). What patterns do you find?

The income variable you want to use is HINCTNT2. To produce (in SPSS version 14.0) a line graph showing several variables, and select the ‘Line Chart’ function from the ‘Graphs’ menu, select ‘multiple’, with data in chart as ‘summaries of separate variables’. The option for confidence intervals (or error bars) is within the ‘Options’ menu

Click here for an example in SPSS syntax.

*Example, looking at self-esteem, autonomy, and trust & belonging, in Germany.

WEIGHT BY DWEIGHT.
 
COMPUTE filter_$=(CNTRY='DE').
VARIABLE LABEL filter_$ "CNTRY='DE' (FILTER)".
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE.
 
MEANS
TABLES=selfesteem autonomy trust_belonging BY hinctnt2
/CELLS MEAN COUNT STDDEV.
 
GRAPH
/LINE(MULTIPLE)=MEAN(selfesteem) MEAN(autonomy) MEAN(trust_belonging) BY hinctnt2
/MISSING=LISTWISE
/INTERVAL CI( 95).
 
FILTER OFF.
USE ALL.
EXECUTE .
Page 4

Distributions of well-being

As with people who look at GDP per capita, there is a tendency for people looking at well-being to focus on the means. Doing so, one completely ignores the distribution. To demonstrate this let’s look at two things.

Exercise 11

Choose an indicator (either an overall index, or one of its constituents). First calculate the standard deviation for that indicator for each country. Which countries have particularly high standard deviations – what does that mean about the distribution of well-being there?

Solution

Let’s take self-esteem. The countries with particularly high variation in self-esteem are France, Hungary and Bulgaria. Those with low variation were Slovakia, Sweden and Cyprus. Note that, in this case, there is little relation between the standard deviation for each country and its mean.

Next, let’s look to see how many individuals in each country suffer ‘well-being poverty’, i.e. having a level of well-being below a certain threshold. Starting from the same indicator, recode it so that individuals with a score of -1 or lower are given a code of 1, and other individuals are coded 0 (label 1 as ‘low well-being’). Next use the crosstabs function to see what percentage of individuals in each country have ‘low well-being’ on that indicator.

Solution

Overall, without the combined weight being applied, 11.1% of respondents had ‘low self-esteem’ (defined as having a score below -1). The countries with the highest proportions suffering low self-esteem were Hungary (18.3%), France (18.2%), and Slovakia (17.4%). The countries with the lowest proportions were Cyprus (4.3%), Germany (5.0%) and Switzerland (5.5%). You need to use the recode function first to produce a new binary variable where ‘1’ means having a self-esteem score below -1 and ‘0’ means having one above -1. Then you need to use the crosstabs function with countries and the new binary variable. Make sure you tick the percentages option such that SPSS calculates percentages within countries. If you are struggling to produce these results, see the SPSS syntax below.

Can you think of a way to visually compare the ranking of countries that this method produces, compared to the ranking produced by just taking the mean well-being for each country? Are there any important differences?

Tip

Why not use the ‘rolling hills’ method, ordering the countries in your chart in terms of the mean of the indicator in question, but only plotting the percentage of people with low well-being? To do this you need to go into Excel. If you do this with self-esteem as described above, you’ll get a graph like the following:

SPSS syntax

* Standard deviations.

WEIGHT BY DWEIGHT.
 
MEANS
TABLES=selfesteem BY cntry
/CELLS MEAN COUNT STDDEV.

* Proportions below a certain level.

RECODE
selfesteem (Lowest thru -1=1) (ELSE=0) INTO low_selfesteem.
EXECUTE.
 
WEIGHT BY DWEIGHT.
 
CROSSTABS
/TABLES=cntry BY low_selfesteem
/FORMAT= AVALUE TABLES
/CELLS= COUNT ROW
/COUNT ROUND CELL.
Go to next chapter >>
Page 5