# Empirical Project 8 Solutions

These are not model answers. They are provided to help students, including those doing the project outside a formal class, to check their progress while working through the questions using the Excel or R walk-throughs. There are also brief notes for the more interpretive questions. Students taking courses using Doing Economics should follow the guidance of their instructors.

## Part 8.1 Cleaning and summarizing the data

• The solution depends on your software and is not shown.
• One way to rename the variables is given in Solution figure 8.1.
Variable New name Variable description
S002EV EVS-wave EVS-wave
S003 Country/region Country/region
S006 Respondent number Original respondent number
S009 Country abbreviation Country abbreviation
A009 Health State of health (subjective)
A170 Life satisfaction Satisfaction with your life
C036 Work Q1 To develop talents you need to have a job
C037 Work Q2 Humiliating to receive money without having to work for it
C038 Work Q3 People who don’t work become lazy
C039 Work Q4 Work is a duty towards society
C041 Work Q5 Work should come first even if it means less spare time
X001 Sex Sex
X003 Age Age
X007 Marital status Marital status
X011_01 Number of children How many children you have—deceased children not included
X025A Education Educational level respondent: ISCED-code one digit
X028 Employment Employment status
X047D Monthly household income Monthly household income (× 1,000), corrected for ppp in euros

Completed data dictionary.

Solution figure 8.1 Completed data dictionary.

1. Examples of answers on how some of the variables were measured are provided:
• To use reported life satisfaction in interpersonal and cross-country comparisons, we need to assume that all individuals share similar presuppositions and answer the questions in similar environments and contexts.

These assumptions are not entirely plausible. Perceived social norms, beliefs, culture, and many other characteristics vary across individuals and across countries, affecting individuals’ responses. As a result, differences in answers may reflect differences in these factors rather than the true level of satisfaction.

There are a large number of environmental factors that can influence the answers. The day of the week, the weather, and major news stories are all examples of such factors. It is impossible to control for all these factors.

Although the measure has limitations, it may be the best we can obtain.

• Employment status: Social norms, beliefs, culture, and other factors which vary across individuals and countries can affect people’s willingness to report truthfully on their employment status. Many people probably do not know accurately the technical definitions of the terms associated with employment status, resulting in misreports. Misreporting of employment status is therefore likely to be an issue.
• An example answer is given below.

• No-opinion responding: A tendency to select the response category that is most neutral in its meaning (for example ‘neither agree nor disagree’). Some people may consider the information being asked for as private and therefore will not want to give opinions. We can check for respondents who choose neutral answers for most questions.
• Primacy effects: A tendency to select one of the first response categories presented on a list. When people want to spend minimal time on a questionnaire, they may just choose the first category given for each question. We can check for respondents who choose the first category for most questions. We can include in the questionnaire similar questions with the categories ordered differently so that if individuals select the first categories even for similar questions, we would know that they are probably not paying attention.
• Socially desirable responding: Conscious or subconscious tendency to select response options more likely to conform with social norms or present the respondent in a good light. The answers reveal the values, attitudes, and beliefs of the respondents. Respondents may engage in socially desirable responding because they are afraid of the risk that their employers or potential employers may learn about their answers and penalize them. We can identify a set of answers complying with social norms in any given region and identify respondents who chose almost the same set of answers.
1. (a)(e) The solution depends on your software and is not shown. For Excel users: Refer to Excel walk-through 8.1 for guidance on answering Question 3(a) and (e).
1. The solution depends on your software and is not shown.
• The solution depends on your software and is not shown.
• The long right tail means the mean is skewed to the right. Using deviations from the average income in this case would underestimate the living standards of many people.
• The solution depends on your software and is not shown. For Excel users: Refer to Excel walk-through 8.3 for guidance, and for an example of naming a new variable.
• There are many possible features to comment on, for example:

• the differences in percentages of students surveyed
• the percentages in full-time employment, part-time employment, and self-employment, making a comparison between developing and developed countries
• variations in unemployment across countries.
Country Full-time Housewife Other Part-time Retired Self-employed Students Unemployed
Albania 29.42 7.42 1.50 5.50 9.08 22.08 7.33 17.67
Armenia 23.86 20.92 1.14 8.09 18.38 5.96 6.70 14.95
Austria 39.80 7.24 1.89 9.95 25.49 5.02 8.39 2.22
Belarus 57.88 2.43 1.21 6.95 18.59 3.40 6.87 2.67
Belgium 42.89 5.96 3.72 8.94 23.01 3.57 5.21 6.70
Bosnia Herzegovina 34.06 9.33 0.82 2.90 14.67 3.08 8.15 26.99
Bulgaria 46.32 2.62 0.76 2.79 31.28 5.58 2.37 8.28
Croatia 41.58 3.37 0.93 2.78 26.01 2.86 8.75 13.72
Cyprus 46.32 13.68 1.29 2.84 24.39 6.58 1.68 3.23
Czech Republic 46.56 3.06 4.66 1.68 31.27 3.82 5.43 3.52
Denmark 55.89 0.28 1.32 6.69 24.32 5.94 4.15 1.41
Estonia 50.35 4.08 2.20 5.11 28.52 3.61 3.38 2.75
Finland 52.34 1.38 3.94 5.11 22.77 6.17 3.72 4.57
France 46.83 5.59 1.94 6.04 28.78 2.76 3.13 4.92
Georgia 19.46 11.60 0.81 6.57 19.38 7.06 2.60 32.52
Germany 38.44 4.58 3.03 8.44 28.64 2.97 2.67 11.23
Great Britain 33.50 7.32 4.01 11.23 28.99 5.72 1.40 7.82
Greece 28.49 17.42 0.40 2.97 26.73 13.72 6.18 4.09
Hungary 46.39 1.20 7.21 2.00 24.04 3.53 6.57 9.05
Iceland 54.50 2.25 6.01 9.91 7.06 11.41 4.95 3.90
Ireland 41.87 19.84 1.59 9.72 13.89 4.96 1.59 6.55
Italy 32.88 8.33 0.46 9.13 23.06 13.70 7.08 5.37
Kosovo 19.57 11.65 0.52 5.23 5.83 9.41 18.00 29.80
Latvia 52.05 6.18 2.34 3.76 23.22 3.26 4.26 4.93
Lithuania 50.13 4.11 2.89 5.16 23.97 3.41 6.04 4.29
Luxembourg 51.33 9.36 1.12 7.38 15.36 3.00 9.87 2.58
Macedonia 35.74 4.34 1.24 1.71 16.74 3.72 8.53 27.98
Malta 33.84 32.33 0.68 3.84 23.42 2.33 0.55 3.01
Moldova 30.49 7.24 1.87 7.58 25.64 4.86 4.43 17.89
Montenegro 39.02 4.63 0.60 2.14 16.64 4.97 4.80 27.19
Netherlands 32.40 9.52 3.68 18.24 27.76 6.48 0.80 1.12
Northern Cyprus 31.19 19.55 2.23 5.20 8.91 8.91 13.61 10.40
Northern Ireland 30.10 10.36 4.53 8.74 29.45 3.56 1.29 11.97
Norway 53.23 2.22 6.55 9.48 12.60 8.17 7.06 0.71
Poland 41.81 6.00 0.10 3.14 28.00 5.81 7.62 7.52
Portugal 46.20 5.24 1.57 3.27 33.51 1.70 1.05 7.46
Romania 41.07 10.54 1.95 3.22 33.95 3.02 3.61 2.63
Russian Federation 54.36 5.81 2.72 5.08 23.77 1.27 2.72 4.26
Serbia 34.21 5.02 1.07 2.38 25.16 6.91 4.11 21.13
Slovakia 40.98 1.73 4.89 2.21 39.73 3.55 1.25 5.66
Slovenia 47.44 2.50 2.75 1.25 31.59 4.37 6.99 3.12
Spain 41.52 16.30 0.11 4.63 19.93 6.28 3.19 8.04
Sweden 54.82 0.38 6.60 7.36 15.36 7.23 4.06 4.19
Switzerland 48.50 6.42 3.21 14.03 21.31 2.89 1.39 2.25
Turkey 16.42 42.39 0.60 2.14 10.00 7.66 5.92 14.88
Ukraine 40.92 6.79 1.02 4.84 32.09 4.41 3.23 6.71

Self-reported employment status in each country (per cent of sample).

Solution figure 8.2 Self-reported employment status in each country (per cent of sample).

• Solution figure 8.3 shows the summary table.
Male Female
Mean Standard deviation Mean Standard deviation
Wellbeing 6.98 2.30 6.98 2.30
Self-reported health 3.68 0.95 3.68 0.95
Work ethic 3.68 0.76 3.68 0.76
Age 47.10 17.42 47.09 17.42
Education 3.09 1.36 3.09 1.36
Number of children 1.63 1.41 1.63 1.41

A summary table for the EVS data.

Solution figure 8.3 A summary table for the EVS data.

## Part 8.2 Visualizing the data

1. Germany is used as an example.
• Solution figures 8.4 and 8.5 provide frequency tables for work ethic in Germany for Wave 3 and Wave 4.
Range of work ethic score Frequency Percentage of individuals (%)
1.00 0 0.00
1.20 3 0.21
1.40 6 0.42
1.60 9 0.63
1.80 15 1.05
2.00 18 1.26
2.20 21 1.47
2.40 47 3.28
2.60 68 4.75
2.80 79 5.52
3.00 114 7.96
3.20 130 9.08
3.40 166 11.59
3.60 171 11.94
3.80 185 12.92
4.00 164 11.45
4.20 106 7.40
4.40 55 3.84
4.60 34 2.37
4.80 20 1.40
5.00 21 1.47

Frequency table for work ethic (Germany, Wave 3).

Solution figure 8.4 Frequency table for work ethic (Germany, Wave 3).

Range of work ethic score Frequency Percentage of individuals (%)
1.0 1 0.06
1.2 1 0.06
1.4 0 0.00
1.6 6 0.36
1.8 9 0.53
2.0 18 1.07
2.2 22 1.31
2.4 35 2.08
2.6 44 2.61
2.8 75 4.46
3.0 90 5.35
3.2 125 7.43
3.4 152 9.03
3.6 171 10.16
3.8 180 10.70
4.0 207 12.30
4.2 191 11.35
4.4 166 9.86
4.6 79 4.69
4.8 37 2.20
5.0 74 4.40

Frequency table for work ethic (Germany, Wave 4).

Solution figure 8.5 Frequency table for work ethic (Germany, Wave 4).

Distribution of work ethic score in Germany: Waves 3 and 4.

Solution figure 8.6 Distribution of work ethic score in Germany: Waves 3 and 4.

• The work ethic scores have increased over time, as shown by the rightward shift in the distribution in Wave 4.
Country Wave 1 Wave 2 Wave 3 Wave 4
Belgium 7.37 7.60 7.42 7.63
Denmark 8.21 8.17 8.31 8.41
France 6.71 6.77 6.98 7.05
Germany 7.22 7.03 7.43 6.77
Iceland 8.05 8.01 8.08 8.07
Italy 6.65 7.30 7.18 7.40
Netherlands 7.75 7.77 7.83 7.99
Northern Ireland 7.66 7.88 8.07 7.82
Spain 6.60 7.15 6.97 7.29
Sweden 8.03 7.99 7.62 7.68

Average wellbeing across countries and survey waves.

Solution figure 8.7 Average wellbeing across countries and survey waves.

• Solution figure 8.8 provides the line chart of average wellbeing across countries and survey waves. Note: When producing a chart with many lines, the key in Excel is often inadequate in allowing the reader to distinguish between the series. You may need to add labels to the series.

Line chart of average wellbeing across countries and survey waves.

Solution figure 8.8 Line chart of average wellbeing across countries and survey waves.

• Average wellbeing increased slightly over the survey periods for most countries. The spread of the distribution of wellbeing will allow us to tell if these changes are statistically significant. Also, instead of looking at averages, we could look at deciles, which would allow us to assess whether the changes in average wellbeing are due to a shift in the entire distribution or an increase in reporting of extreme values.
Variable Wellbeing Work ethic
Age –0.08 0.13
Education 0.09 –0.15
Employment status (= 1 if full-time employed, = 0 if unemployed) 0.18 –0.03
Gender (= 0 if male, = 1 if female) –0.02 –0.05
Self-reported health 0.38 –0.07
Income 0.24 –0.15
Number of children –0.02 0.09
Relative income 0.19 –0.05
Wellbeing 1.00 –0.03
Work ethic –0.03 1.00

Correlation between wellbeing, work ethic and other variables.

Solution figure 8.9 Correlation between wellbeing, work ethic and other variables.

• The correlation coefficient, ranged between –1 and 1, measures the strength and direction of a linear relationship between two variables. A negative coefficient means the two variables are negatively linearly correlated, and a positive coefficient means the two variables are positively linearly correlated. The coefficient of 0.18 between employ­ment status and wellbeing, for example, means that people who are full-time employed tend to be more satisfied with their lives relative to those who are unemployed. The coefficients between wellbeing and other variables have the expected sign. The results that work ethic is negatively linearly related to education, employment status, and relative income, however, are somewhat surprising.
• Solution figure 8.10 shows average wellbeing according to employment status.

Employed respondents are more satisfied with their lives compared with the unemployed. The retired, like the unemployed, do not work, but they do not suffer nearly as large a drop in life satisfaction. This suggests that social norms that discriminate against the unemployed is a plausible explanation for the relatively low wellbeing of the unemployed.

Average wellbeing Employment status
Country/Region Full-time Retired Unemployed
Albania 6.63 5.81 6.07
Armenia 6.04 4.85 5.46
Austria 7.44 7.74 6.07
Belarus 6.10 5.62 5.61
Belgium 7.72 7.83 6.37
Bosnia Herzegovina 7.33 7.01 6.77
Bulgaria 6.18 4.97 4.69
Croatia 7.31 6.48 7.17
Cyprus 7.38 7.03 6.56
Czech Republic 7.30 6.89 6.07
Denmark 8.54 8.21 7.20
Estonia 6.93 6.25 4.97
Finland 7.82 8.02 5.77
France 7.20 6.97 6.24
Georgia 6.12 4.69 5.42
Germany 7.26 6.85 4.61
Great Britain 7.53 7.93 6.03
Greece 7.15 6.62 5.98
Hungary 6.65 5.89 4.86
Iceland 8.20 8.45 7.23
Ireland 7.90 7.83 7.18
Italy 7.43 7.44 6.60
Kosovo 6.30 6.04 6.78
Latvia 6.52 5.93 5.31
Lithuania 6.59 5.63 4.53
Luxembourg 7.87 8.24 5.50
Macedonia 7.19 6.67 6.61
Malta 7.70 7.76 5.95
Moldova 7.12 5.98 6.07
Montenegro 7.63 7.22 7.47
Netherlands 8.04 7.95 7.00
Northern Cyprus 6.74 6.44 5.64
Northern Ireland 7.68 7.77 7.54
Norway 8.19 8.26 8.00
Poland 7.46 6.57 7.06
Portugal 6.84 5.88 5.44
Romania 7.14 6.57 7.41
Russian Federation 6.86 5.70 6.40
Serbia 7.17 6.67 6.73
Slovakia 7.47 6.71 6.12
Slovenia 7.83 7.13 6.76
Spain 7.34 7.21 7.19
Sweden 7.88 8.17 6.52
Switzerland 8.04 8.12 5.76
Turkey 6.50 6.61 5.76
Ukraine 6.34 5.44 4.95

Average wellbeing according to employment status and country.

Solution figure 8.10 Average wellbeing according to employment status and country.

Country/Region Difference between full-time employed and unemployed Difference between full-time employed and retired
Albania 0.57 0.83
Armenia 0.58 1.18
Austria 1.36 –0.31
Belarus 0.49 0.48
Belgium 1.35 –0.12
Bosnia Herzegovina 0.56 0.33
Bulgaria 1.48 1.20
Croatia 0.14 0.83
Cyprus 0.82 0.35
Czech Republic 1.24 0.42
Denmark 1.34 0.34
Estonia 1.95 0.67
Finland 2.05 –0.20
France 0.96 0.23
Georgia 0.70 1.43
Germany 2.65 0.41
Great Britain 1.51 –0.40
Greece 1.17 0.53
Hungary 1.79 0.76
Iceland 0.97 –0.24
Ireland 0.71 0.07
Italy 0.83 –0.01
Kosovo –0.49 0.26
Latvia 1.21 0.59
Lithuania 2.06 0.96
Luxembourg 2.37 –0.37
Macedonia 0.58 0.52
Malta 1.75 –0.06
Moldova 1.05 1.14
Montenegro 0.16 0.42
Netherlands 1.04 0.09
Northern Cyprus 1.10 0.29
Northern Ireland 0.14 –0.09
Norway 0.19 –0.07
Poland 0.40 0.89
Portugal 1.40 0.96
Romania –0.27 0.56
Russian Federation 0.46 1.16
Serbia 0.44 0.50
Slovakia 1.35 0.76
Slovenia 1.07 0.70
Spain 0.15 0.13
Sweden 1.36 –0.30
Switzerland 2.28 –0.08
Turkey 0.74 –0.11
Ukraine 1.40 0.90

Difference in average wellbeing: full-time employed minus unemployed, and full-time employed minus retired.

Solution figure 8.11 Difference in average wellbeing: full-time employed minus unemployed, and full-time employed minus retired.

• Solution figures 8.12 and 8.13 provide column charts showing both differences in wellbeing.

Difference in wellbeing between the full-time employed and the unemployed (sorted from lowest to highest average work ethic).

Solution figure 8.12 Difference in wellbeing between the full-time employed and the unemployed (sorted from lowest to highest average work ethic).

Difference in wellbeing between the full-time employed and the retired (sorted from lowest to highest average work ethic).

Solution figure 8.13 Difference in wellbeing between the full-time employed and the retired (sorted from lowest to highest average work ethic).

• There is no clear relation between work ethic and the difference in wellbeing: if there were, the bars would be increasing (or decreasing) in height as we moved to the right along the horizontal axis.

## Part 8.3 Confidence intervals for difference in the mean

1. Choice of countries for this example:

• top third: Turkey
• middle third: Spain
• lower third: Great Britain.
• The table for our chosen countries is as shown below.
Full-time Retired Unemployed
Country Average SD Count Average SD Count Average SD Count
Great Britain 7.53 1.85 334 7.93 1.98 289 6.03 2.19 78
Spain 7.34 1.70 377 7.21 1.93 181 7.19 2.14 73
Turkey 6.50 2.55 330 6.61 2.60 201 5.76 3.13 299

Summary table of wellbeing, by employment status.

Solution figure 8.14 Summary table of wellbeing, by employment status.

• The calculations are provided in Solution figures 8.15 and 8.16.
Country Difference in means SD of difference in means Number of observations
Great Britain –0.40 2.71 623
Spain 0.13 2.58 558
Turkey –0.11 3.64 531

Calculated values for differences in wellbeing (full-time vs retired).

Solution figure 8.15 Calculated values for differences in wellbeing (full-time vs retired).

Country Difference in means SD of difference in means Number of observations
Great Britain 1.51 2.86 412
Spain 0.15 2.73 450
Turkey 0.74 4.04 629

Calculated values for differences in wellbeing (full-time vs unemployed).

Solution figure 8.16 Calculated values for differences in wellbeing (full-time vs unemployed).

• The calculations are provided in Solution figures 8.17 and 8.18.
Country Difference in means SD of difference in means Number of observations CI distance to sample mean Total width of CI
Great Britain –0.40 2.71 623.00 0.21 0.43
Spain 0.13 2.58 558.00 0.21 0.43
Turkey –0.11 3.64 531.00 0.31 0.62

Calculated width of 95% confidence interval for differences in wellbeing (full-time vs retired).

Solution figure 8.17 Calculated width of 95% confidence interval for differences in wellbeing (full-time vs retired).

Country Difference in means SD of difference in means Number of observations CI distance to mean Total width of CI
Great Britain 1.51 2.86 412.00 0.28 0.55
Spain 0.15 2.73 450.00 0.25 0.51
Turkey 0.74 4.04 629.00 0.32 0.63

Calculated width of 95% confidence interval for differences in wellbeing (full-time vs unemployed).

Solution figure 8.18 Calculated width of 95% confidence interval for differences in wellbeing (full-time vs unemployed).

• Solution figures 8.19 and 8.20 provide column charts for the example countries showing the difference in wellbeing and confidence intervals.

Difference in wellbeing (full-time and retired).

Solution figure 8.19 Difference in wellbeing (full-time and retired).

Difference in wellbeing (full-time and unemployed).

Solution figure 8.20 Difference in wellbeing (full-time and unemployed).

• The difference in means is statistically significant if the confidence interval for the difference in means does not contain 0. The differences in means are significant in Great Britain. The difference in means between the full-time employed and the unemployed in Turkey is also significant. The rest of the differences are not statistically significant at a 5% significance level.
• A natural experiment can control for variables that affect both the dependent and independent variables. The method allows us to isolate the effect of employment status on wellbeing. The ceteris paribus effect of employment status on wellbeing, if obtained, has a causal interpretation.
• If we have data on the same individual over time, then we can calculate the difference in wellbeing for the same person under different employment statuses, instead of comparing different individuals (as we did in this project).