Analysis on the Effect of Education Level and Career Choice on Happiness An Empirical Research Using Open Data
Main Investigator: Vincent van der Holst Student number: 10088385
First Supervisor: Vladimer Kobayashi Second Supervisor: Dr. Gábor Kismihók
Disclaimer
This paper uses data from SHARE wave 4 release 1.1.1, as of March 28th 2013 or SHARe wave 1 and 2 release 2.6.0, as of November 29th 2013 or SHARELIFE release 1, as of November 24th 2010. The SHARE data collection has been primarily funded by the European Commission through the 5th Framework Programme (project QLK6-CT-2001-00360 in the thematic programme Quality of Life), through the 6th Framework Programme (projects SHARE-I3, RII-CT-2006-062193, COMPARE, CIT5-CT-2005-028857, and SHARELIFE, CIT4-CT-2006-028812) and through the 7th Framework Programme (SHARE-PREP, N° 211909, SHARE-LEAP, N° 227822 and SHARE M4, N° 261982). Additional funding from the U.S. National Institute on Aging (U01 AG09740-13S2, P01 AG005842, P01 AG08291, P30 AG12815, R21 AG025169, Y1-AG-4553-01, IAG BSR06-11 and OGHA 04-064) and the German Ministry of Education and Research as well as from various national sources is gratefully acknowledged (see www.share-project.org for a full list of funding institutions).”
Abstract
The goal of this paper was to do exploratory research in open data on labour market
characteristics. Data from SHARE was chosen. Using the open data from SHARE, a labour market analysis of the effects of educational attainment, career choices and an interaction term between education and career choice on happiness was done. Correlation analysis found a positive relationship between education and happiness, in agreement with earlier research. Career choice effect on happiness depends on the type of job industry. Most job industries were found to positively affect happiness and one was found to negatively affect happiness. The interaction of education and career choice produced mixed results, mostly small negative and positive effects were found and most of the combinations were significant, where other interactions were found to be insignificant. These findings have implications for job
applicants and students as well as managerial implications. They benefit from the added knowledge on variables influencing quality of life, especially from career choice affecting happiness, which previously was a gap in research.
Keywords: open data, database, labour market, happiness, quality of life, education, data mining and big data.
Acknowledgements
First and foremost, I would like to thank my thesis supervisor Vladimer Kobayashi for his tremendous help during all parts of this research. Tapping from his knowledge on statistics helped me improve results and findings. SHARE and its contributors are thanked for making their databases openly available and providing easy access and clear instructions on data processing. The knowledge on labour market characteristics, education, happiness and career choices made available by previous research was a crucial element this thesis builds upon. Therefore, the authors of the articles I referred to are thanked for their contribution to this thesis.
Table of contents Page number Disclaimer Abstract Acknowledgements 1. Introduction………1
1.1 Background of the Study………...1
1.2 Statement of the problem………...1
1.3Goals and Objectives………..2
1.4Scope and Limitations………..……..3
2. Literature review……….………...4 2.1 Existing literature………..…..……….. 5 2.2 Education……….…….……….6 2.3 Happiness………..……..………..7 2.4 Career Choice………...…..………...8 3. Conceptual framework……….……..………..8 4. Methodology………...………9 4.1 Databases………...………9 4.1.1 Share………..……….10 4.1.2 Eurostat……….……….10 4.1.3 Quandl……….…………...10 4.1.4 Laborsta……….……….11 4.1.5 Ilostat……….……….11 4.2 Method.……….11
5. Results and discussion.……….………13
5.1 Assumptions……….…………14 5.2 Summary statistics………17 5.3 Anova’s ……….21 5.4 Regression analyses………..……….22 5.4.1 Regression Anova’s……….………22 5.4.2 Regression analyses….……….……..24
5.4.3 Comparing the education and happiness relationship to existing literature………...………..24
5.4.4 Regression of job industries and quality of life………….….25
5.4.5 Comparing the career choice and happiness relationship to existing literature……….…..….26
5.5 Interaction effects……….…….27
5.6 Theoretical implications……….…………...30
5.7 Practical implications………31
6. Conclusion………..………32
7. Recommendations for future work………..33
7.1 Suggestions for future research……….……….33
Bibliography………..………..35
1. Introduction 1.1 Background of the study
Individuals pursue higher education to acquire adequate knowledge for the dream job they always wanted. Motivational factors for studying are intrinsic and extrinsic in nature. Intrinsic motivations for studying are to know, to accomplish and to experience (Vallerand, Pelletier, Blais, Briere, Senecal & Vallieres, 1992). A qualitative study done by Henderson (2000) found that individuals who follow career paths based on their interests, what they succeeded in and enjoyed are happier than those who do not. Career indecision is one of the most frequent problems college students have (Osipow, 1999). So career choice seems to be an important factor in the future of individuals and should not be neglected when researching factors influencing happiness. Sixty-nine percent of people surveyed by Diener (2000) rate happiness highest on the importance scale, so it is clearly important to follow a life path that potentially leads to happiness. Many researchers (Lent, Brown & Hackett, 2000; Henderson, 2000) stress the importance of careers in individual’s happiness, but they do not come with data backing these claims and actually helping people understand the career factors that influence their quality of life. This research hopes to address at least one of these issues by taking a look at career choice effects on quality of life1.
In order to fill the existing research gap on career choices influencing the relationship between education and happiness and add to the readily available literature a research
question has to be formulated, which will be done in the subsequent goals and objectives section. Before researching career choice effect on happiness, the relationship between education and happiness will be tested.
1.2 Statement of the Problem
Given the levels of education and the type of job industry, there should be clear predictions as to what combinations lead to the highest level of happiness. Primary level education and a job in mining might lead to high levels of happiness, whereas a tertiary level of education and a career in mining could result in individuals being unhappy in their current situation. Providing this information and making it accessible to the public will yield individuals with the
knowledge to make a fitting career choice given their level of education, possibly resulting in maximization of their happiness.
As stated above, knowing is a major motivational factor for studying (Vallerand et al., 1992). However, current research relies heavily on the single relationship between education,
1 Quality of life and happiness are interchangeably used in this study
1
in both years and level, and the relation these have on happiness (Noval and Garvi, accessed July 11, 2014; Powdthavee, Lekfuangfu & Wooden 2013; Hartog and Oosterbeek 1998). But does higher education actually lead to happiness? This research hopes to, among others, answer this question.
In spite of the gap in current literature, opportunities for expanding and complementing existing knowledge are possible. Education is just one of the factors influencing happiness. Crucial to future research is to increase variables that might relate to happiness. By adding moderating variables such as career choice to the relation between education and happiness, the knowledge gap will diminish.
1.3 Goals and Objectives
One of the main goals of this research was to do an exploratory search for open datasets in education and the labour market. A great amount of useful data has already been made openly available. Data across cultures, countries and even continents are only mouse-clicks away. Researchers have been eager to use these massive and unrestricted datasets, which produced numerous studies that generated new insights into education and the labour market. Big enterprises also try to gain an edge on their competitors by mining data and streamlining there processes and services to the data characteristics.
Despite these recent efforts towards gathering data and conducting research with these databases, there is still room for contribution. Research has not yet addressed an existing gap in research, although data that can bridge this gap is openly available. One improvement in current research is the before mentioned “career choice” as a factor influencing the
relationship between education and happiness. Naturally, a main research question and some hypotheses have been constructed to try and bridge this literature gap.
Consequently, the main research question is: Does career choice moderate the positive relationship between education and happiness?
To aid in the answering of this question a couple of additional hypotheses are
formulated. First of all, the data will show that, in accordance with previous research, there is a positive relationship between education and happiness.
Hypothesis 1 states: Level of Education is positively correlated to level of happiness. Hypothesis 1a: Happiness increases for every incremental level of education
Secondly, analysis will show if there is a relationship between career choice and happiness. Hypothesis 2: Career choices affect happiness.
Finally, the third hypothesis will show if career choice moderates the relationship between education and happiness: There is an interaction effect between level of education and career
choices on happiness. These hypotheses will lead to a more unified theory on both the
implications of education on the labour market and as well the influence career choice has on happiness. The interaction term between education and career choice will finally lead to a better understanding of the relations of career choice and education combinations have on happiness.
To answer the main research question and corresponding hypotheses, an open data analysis with the relevant variables will be done. The outcome of this analysis will render some insights into the relationships between education, career choice and happiness and might prove to be useful for future reference by other researchers looking to add to the current research in this field. Arguably more important are the effects these results might have on students. The outcomes can benefit students looking to find a fitting study or graduates searching for employment in a particular industry. Graduates might find the results beneficial to choose the direction of their career paths. Recent research by Boehm & Lyubomirsky (2008) unveils another contribution this research could have to society. It shows that happy persons outperform their less happy peers in terms of salary, performance and helpful act. These outcomes suggest that it is important for companies to have happy employees on their payroll to improve company performance. They should make efforts to put the right people in the right place, something that might be accomplished by using the results from this empirical study.
1.4 Scope and limitations
Limited time and resources restrict what research can accomplish. Adding time, results can be corrected for external influences that have an effect on the researched variables. For example, happiness is influenced by a lot more than solely career choices and educational attainment. Health, environment and family could possibly change outcomes when they are accounted for. Even though the data is inclusive of these variables, time limitations restricted the adjustments for most externalities. Future research is encouraged to expand this work by adding factors influencing the relationships in this study.
Like every research, this paper will have its limitations. There is no such thing as perfect data. Open data strives to eliminate many of the shortcomings that small surveys and qualitative analysis have. The substantial number of respondents improves reliability and generalizability. Still, only part of the wanted information can be retrieved, no data possesses the ability to include every variable from all regions, and therefore no sample perfectly resembles the population. Even for data across many nations and cultures generalizability is an issue, some cultures do not match others, and small differences will always be present.
Inhabitants from different cultures might perceive questions from surveys in different ways, thus changing outcomes.
Gathering data in the developed world is relatively easy due to population
demographics and the advanced internet infrastructure. Big enterprises push the gathering of data to ever improve their services, so they match their customers. In developing countries it is more difficult to tap and synthesize data by cause of the lack of reliable infrastructure and the low availability of necessary expertise (Anokwa, Hartung, Brunette, Borriello & Lerer, 2009). The resulting limitation is that data on developing countries is far less available compared to data from the developing world. The transformation the internet generation established will close this gap. Even business in Africa is adopting internet services more and more, even though it still trails behind other continents (McKinsey, 2014). This study also lacks data on the developing world, resulting in a limitation on geographical level.
A limitation more specific to this study is response bias. The majority of data is collected through surveys, which might include central tendency bias and the halo effect (psychology lexicon, 2014). Assumptions about data might not always be correct and produce faulty results. Outcomes might be different across continents and even regions within
countries.
Following this introduction, this paper will continue with a literature review that outlines previous research relevant to this study. Literature will clarify definitions and previous research on the subject. A conceptual model will visualize the research design and methodology, respectively the next topics in this paper. The results section displays the outcome of the various statistical analyses. Subsequently, the discussion section will provide the theoretical and practical implications of the findings and suggestions for future research this study bears with it. This paper ends with some concluding remarks.
2. Literature review
In the literature section, existing literature that relates to the research question and hypotheses will be reviewed to get a clear understanding of the different definitions of the research. Past literature on the subject will give insight in the current status of this research field and clarify in which departments it is lacking. This research focuses on the relationship between
education and happiness, moderated by career choices. First of all, it is crucial for this research to provide clear definitions of education, happiness and career choices to make sure there is no misunderstanding of the definitions of these variables. Furthermore, existing models including relationships between the variables included in this research will be
reviewed to acquire a understanding of the current knowledge on the subject. Some concluding remarks on the literature give a short oversight before carrying on with the conceptual model.
2.1 Existing literature
Extensive research has been done on the relation between education and happiness (Noval and Garvi, accessed July 10, 2014; Powdthavee, Lekfuangfu & Wooden 2013). Hartog and
Oosterbeek (1998) find that intermediate level students rate the highest on a happiness scale, they also find that women are on average happier than men, even though they earn less. The study also found that a higher IQ does not affect wealth and happiness. In a study using Swedish microdata, Gerdtham and Johannesson (2001) find that Happiness increases with education. Michalos (2008) finds a small significant relation when looking at the relation between education and happiness, but when adding informal education to the equation there is a more substantive positive relationship between education and happiness. A Harvard study (Fowler, Christakis, Steptoe & Roux, 2009, p.2) finds that educated people are slightly happier. According to Veenhoven (1996) a small and variable effect of gender is noticed. A general consensus that higher levels of education positively correlate to happiness seems to emerge, although research definitely is not unanimous on this fact. Most studies find the relationship between education and happiness to be significant but small. Therefore, this study aims to find a relationship that significantly differs from past research by adding a moderating variable to the relationship between education and happiness, in this case career choices. It is important to note the distinct difference between the effects of education and income on happiness. Education is consistently found to positively correlate with income, a study done by Easterlin (1995) that summarizes 45 happiness surveys find that increasing the income of a whole country will not result in an increase in happiness. Thus, education increases income and happiness, but an increase in income does not necessarily result in an increase in happiness. Thus it is important to keep possible causal relationships in mind. A study by Gregorio & Lee (2002) found that increasing average educational attainment results in a more equal income distribution throughout a country. Studies on the effects of career choices on happiness are significantly fewer. Henderson (2000) found that people who follow their instincts achieve career happiness, which leaves it open for discussion if career happiness results in overall happiness. Being a qualitative study with only 8 participants, reliability is another issue. Lent Brown & Hackett (2000) list study barriers as one of the suggestions to why people choose whether to pursue a particular career option. Education could be a
necessary step to gain access to certain jobs. Higher levels of education may lead to more options in more industry sectors.
This research will attempt to, yet again, find evidence for the relationship between educational attainment and happiness. The other independent variable: job industry, also referred to as career choice, will be analysed to see if it has an effect on happiness. In addition, career choices as a moderating factor in the relationship between education and happiness will be analysed for possible effects.
2.2 Education
This research will define education in accordance to the ISCED (INTERNATIONAL STANDARD CLASSIFICATION OF EDUCATION) coding (ISCED, 2014). The ISCED coding identifies 6 different levels of education, originating from the 3 levels of formal education: primary, secondary and tertiary education. ISCED coding is consistent across cultures and demographics, it is therefore especially suited for research that addresses multiple countries. Most research codes education on different levels, and many use the ISCED coding. Comparability between this and other research making use of the ISCED is therefore easier. One could also use non-formal forms of education, learning from the
environment, work, news and personal experiences. This broadens the definition of education and therefore makes it harder to define. In addition, informal education is difficult to quantify. Consequently this research will focus on the formal levels of education, increasing the odds of finding consistent and valuable data. Table 1 shows the different definitions for each level of education.
Table 1: ISCED scale describing the six levels of educational attainment Description
1 Primary
2 Lower secondary
3 Upper-secondary
4 Post-secondary non-tertiary 5 First stage of tertiary 6 Second stage of tertiary
2.3 Happiness
Happiness or more generally termed: quality of life, is more difficult to define than education. Reversely coded scales for happiness exist, for example a depression scale, where a higher rating corresponds to lower happiness. The happiness variable often stems from multiple scaled items and is then transformed into a single “happiness” variable. A study done by Easterlin (2001) strives towards a unified theory for income and more importantly, happiness. This paper and its references note that happiness is subjective and comparability issues arise as a result. However, Cantril (1965) found that across cultures and widely different stages of socio-economic development people mention material circumstances most often when asked what they want out of life. Family concerns, personal or family health and matters relating to work are named in decreasing order as concerns. As happiness is ordinarily determined by factors such as family, health and living, these results from 14 countries and over 20000 respondents suggest that, although everyone is free to define happiness in his or her own way, the kind of things that shape happiness are similar for most people. These findings improve comparability and decrease subjectivity. A study conducted by Shin and Johnson (1977) on happiness found that it is primarily a product of the positive assessments of life situations and favourable comparisons of these life situations with those of others and in the past.
Rights, gender equality and individualism are among the variables that positively correlate with happiness indicators (Schyns, 1998). Among elderly people, health and social contact are rated as one of the highest factors affecting happiness (Wiggins, Higgs, Hyde & Blane, 2004). An extensive research on the causes and correlates of happiness conducted by Argyle in 2003 concludes that around 15% of the variance of happiness is caused by
demographics, with most factors having only a slight effect. Personal factors contribute much more to the variance of happiness, with marriage being the highest correlate. Married people are the happiest. Unemployment is another major factor causing people to feel unhappy. Retirement is also found to benefit happiness. Other factors having positive effect on
happiness include sport and exercise, social clubs, music and voluntary work. Religion has a small positive effect. After correcting for income, education and occupation, ethnic
minorities are only slightly less happy. To summarize, there are many factors found to correlate with happiness, some more than others. Comparability appears not to be an issue, because of equal perceptions of happiness among demographics.
This study will use the CASP-12 (Hyde, Wiggins Higgs & Blane, 2003) score that measures quality of life and is based on four subscales on control, autonomy, pleasure and self-realization (Guide to EasySHARE, 2013). The quality of life measure was developed for
older people as a self-completion questionnaire based on Likert scaled items. Examples of the scale include “My health stops me from doing the things I want” and “Shortage of money stops me from doing the things I want to do” (Howel, 2012). The CASP-12 scale ranges from 12 to 48, minimum happiness is 12 and maximum happiness results in a score of 48.
2.4 Career choices
Career choice is the third and final variable relevant to this research. In this study the NACE, an acronym of the General Industrial Classification of Economic Activities within the
European Communities, will be used (1990). Over the years, countries have either developed their own classifications or used international systems to meet their needs. The result of such an uncoordinated approach is a series of national systems which, although they may meet national needs, are unsatisfactory for making international comparisons, let alone for operating a single market. Different types of jobs in NACE are ranked in 13 corresponding industries like mining, manufacturing and construction (NACE codes, 1993). The NACE scaling makes it easier to interpret results because it has been extensively used in past research and can therefore be compared to past research (Kromhout, 2003; Bednarzik, 2000; Mroczek & Kolarz, 1998).
3. Conceptual framework
A conceptual framework is formulated to visualize the various parts forming this research. It harbours the key variables and hypotheses of the study. The conceptual framework provides a brief overview of the independent and dependent variables. Figure 1 shows the conceptual framework.
Figure 1: Conceptual framework with education and career choices as independent variables and happiness as dependent variable.
As shown by Figure 1, this study has 2 main independent variables, namely “education” and “career choices”. The impact of these 2 independent variables on the dependent variable happiness will be researched, making use of this framework. The variable Education is based on ISCED-97 (ISCED, 2014) scaling, which incorporates 6 different levels of education. The NACE (NACE, 2014) coding for the different job industries is a concise measurement scale to group jobs into their corresponding industries, in this case 13 industries and “other”. Level of education is hypothesized to have a positive effect on happiness, with higher education level resulting in a higher happiness rating. The categorical variable career choice is proposed to have a significant effect on happiness as well. Finally, the interaction term Education*career-choices is hypothesized to have a relationship with happiness. Happiness is rated according to the CASP-12 scale (Hyde et al., 2003).
4. Methodology 4.1 Databases
Before the final databases were chosen, an exploratory search for data began. Multiple data sources were considered. Important factors in the search for data are the size; bigger datasets are favoured in this research because they are more likely to produce significant results and they better represent the population. Specifically, data across regions and containing multiple demographics were preferred. Widespread data sources improve comparability, outcomes can be compared to different demographics. Another factor that
simplifies data comparability is if it contains scales and coding that is widely used in research. Therefore, it must not be overlooked how the variables are coded. Because this research’s scope was to find data about labour markets, education and career choices, keywords are framed to streamline the search for data. Keyword examples are: open data, database, labour market, happiness, quality of life, education, data mining and big data. Weighing the
mentioned factors and putting databases side by side, the most optimal database for this research was found. Below there is a list of databases that were considered before the most fitting one was chosen.
4.1.1 SHARE
The Survey of Health, Ageing and Retirement in Europe (and Israel) is a multidisciplinary and cross-national panel database of micro data on health, socio-economic status and social and family networks of more than 85,000 individuals (SHARE, 2014). This database contains data from respondents aged 50 years and up. Variables include health statistics such as BMI and blood sugar, labour market information about job industries and professions, income and demographics. Access has to be attained by filling out a form that states your purposes and carefulness concerning the data (Share statement, 2014).
The data from SHARE is grouped in 4 different “waves”. Every wave consists of multiple datasets. Apart from the SHARE wave there is also SHARELIFE: The questionnaire of SHARELIFE has a different focus than the regular waves. It contains all important areas of the respondents’ live histories, ranging from childhood conditions, partners and children, housing and work history to detailed questions on health and health care (SHARE, 2014). For education share uses the ISCED (INTERNATIONAL STANDARD CLASSIFICATION OF EDUCATION) coding (ISCED, 2014). Different types of jobs are ranked in different
industries like mining, manufacturing and construction. Industries are coded according to the NACE standard (NACE, 1993). The CASP-12 scale is used to quantify quality of life,
4.1.2 Eurostat
Eurostat is the statistical office of the European Union situated in Luxembourg. Its task is to provide the European Union with statistics at European level that enable comparisons between countries and regions (Eurostat, 2014). Data from Eurostat ranges from labour market statistics such as income and employment factors. On a national level it provides information about GDP, growth rates, bond yields and government deficits.
4.1.3 QUANDL
Quandl is a dataset search engine that offers free and unlimited access to 8 million time-series datasets from 400 sources spanning finance, economics, society, health, energy, demography
& more (Quandl, 2014). The Quandl website aims to simplify the data search and it was used as one of the means to search for fitting databases in this research.
4.1.4 Laborsta
Laborsta covers official core labour statistics for over 200 countries from 1969 to 2008. This database is not being updated and most series were migrated to ILOSTAT (Laborsta, 2014).
4.1.5 Ilostat
Just like the Laborsta database, this new ILO database of labour statistics provides multiple datasets with annual and infra-annual labour market statistics for over 100 indicators and 230 countries, areas and territories (ILOSTAT, 2014). Even more so than the other databases, the widespread sources of data make this a powerful tool to perform research on the labour market.
After careful consideration, the SHARE database was chosen. There are several underlying reasons for this choice. Firstly, because the data covers so many variables, most of which are interlinked more or less with this research, no compromises have to be made on the proposed research. Data is on a personal level, where other databases such as Eurostat and Ilostat use data on national- and regional-level. This makes it possible to do a case-by-case comparison of the data. The cross-national nature of the data aids comparability across cultures and demographics.
4.2 Methods
After the databases from SHARE are chosen, the relevant datasets were exported to excel. In excel preparations will be made so the data is complete and has no errors. Then the data is exported to SPSS where analyses will commence.
More specifically this research will use data from SHARE (2014), SHARElife (2014) and EasySHARE (2014). The SHARE job episodes panel data (2013) contains, among other variables, data across 3 waves of data collection about the labour market history of the
respondents. Job data is coded according to the General Industrial Classification of Economic Activities within the European Communities, acronymed as NACE (1990). Because this study will require multiple datasets in order to acquire all the needed variables it is necessary to merge different datasets. Fortunately SHARE provides a so called “mergeid” that is unique and non-changing for every respondent in the data, across all the different datasets and waves of data. An example of a mergeid is DK-317238-02, the letters stand for the country code (Denmark in this case), the middle is a unique code for every participant, and the latter 2 numbers represent the wave the data originates from (SHARElife release guide, 2010). With this unique id it is possible to merge multiple datasets and construct one dataset that contains
all the needed variables. SPSS deals with the merging of the, in this case, two datasets. When the datasets are merged, variables can be constructed to test the hypothesis that career choices moderate the relationship between education and happiness. Because the dataset contains a plethora of variables it can be possible to include more variables. The original “job episodes” database contains information for every year of a respondent's past, this resulted in multiple identical cases for one respondent. To correct the data to only include one case per respondent, data is filtered by age, and because the SHARE data only includes respondents aged 50 years and up, the decision is made to filter data to only include the age 50. Therefore, the job industry the respondent was working in at age 50 is used, as opposed to the multiple job industries a respondent might have worked in during their active career. Education is split into the different levels of education (6 levels). “Still in school” and “none” are the other available options; they are however coded out so only educated respondents are included in this research, to stress the focus on contributing to students and businesses. Career choice will be split according to the before mentioned NACE coding scale, which ranks jobs into 13 industry categories and “other”, which will be ignored in the data, because it is unclear what these jobs encompass. The SHARE databases rank missing, incomplete or false entries as negative values, those have to be excluded first, so only complete and usable data is left. Before starting the analysis, it is critical to check assumptions about the predictor variables. Linearity can be checked by drawing a scatter plot. As a final check for assumptions,
multicollinearity has to be ruled out using the collinearity diagnostics in SPSS. Starting of the analysis will be some descriptives of the variables, giving an idea of the mean level of
education and happiness. Frequencies of the different categories: job industry, education level, gender and level of happiness give insight in the numbers involved. Moving on with statistical methods it has to be noted that education is categorical, but contestably has equal steps between the levels of education and therefore education will be correlated with
happiness to check if there is correlation between the two. After these analyses, education will be dummy coded, with the lowest level of education being the reference group. With the dummy coded variables ready, regression will render the linear association between the levels of education and happiness. Secondly, career choices will be analysed nearly identical to education. Because industry is categorical and industries do not represent equal steps, job industry needs to be recoded due to the categorical nature of the variable. With the use of dummy variables the different job industries will be coded. Again, the industry categories will be compared against each other, proving their effects on happiness relative to each other. Regressing the dummy coded industry variable onto happiness will yield the separate linear
associations between the variables. To test the main research question, namely the moderation effect, the interaction term education*career-choice was added. To make the interaction term, data was again exported and then imported in the statistical program called R. Producing the interaction term is done by multiplying the dummy coded job industries with the levels of education, also coded as dummy variables. The independent variable education*career-choice and the dependent variable happiness are put through regression analysis to reflect the
directions of these relationships. Formula 1 is produced to visualize the mathematics behind the interaction terms. It depicts the average means of happiness for all the combinations of educational attainment and career choice.
Formula 1: Interaction formula for the moderating effect of career choice on the educational attainment and happiness relationship.
y = β0 + β1+ β2 + β3 + ε Where:
y=the average of happiness β0= the intercept
β1= education beta. β2= career choice beta
β3= beta of interaction term education*career choice ε=the error term
Using this formula average means for the educational attainment and career choice combinations on happiness can be calculated.
The outcome of these analyses will show if careers in different industries moderate the effect of education on happiness and thus, if the main research question is supported. The results and discussion section following this methodology will give insights into the outcomes of the statistical analyses described here.
5. Results and discussion
This section will follow the methodology, using its steps to come with useful results and explaining and elaborating the findings these results imply. Firstly, the underlying assumptions will be checked, multicollinearity diagnostics are used to check for correlation among the predictor variables. Statistics summaries will provide a view on the numbers at hand. Bar charts and frequencies are inclusive of the statistics (They are appended). Running both one-way Analysis of variances (ANOVA) and regression ANOVA’s to establish
differences between groups were the subsequent steps in the statistical analyses. The first regression between the dummy coded educational attainment (IV) and happiness (DV) is described. The second regression, between the other independent variable career choice and the dependent variable happiness was done. Finally, all the interaction terms education*career are added in the regression model, again with happiness as the dependent variable. Outcomes are checked with the information from the literature review to see if they correspond or
contradict each other. After this, the outcomes of these regressions will either support or reject the proposed hypotheses in this study. Succeeding the statistical analysis, this part will list some theoretical and practical implications that this study bears with it.
5.1 Assumptions
Figure 2: Boxplot of educational attainment (ISCED) and happiness (CASP)
Figure 3: Boxplot of job industry and happiness (CASP)
Taking a look at the boxplots, for both the relationships between the dependent and the independent variables there appears to be a relationship between the variables. Next, testing for multicollinearity is done to check for strong relationships among the independent variables, an undesired attribute of linear regression.
Table 2: Multicollinearity diagnostics for education attainment levels:
Education level VIF
edu level 1 4,116 edu level 2 3,355 edu level 3 4,321 edu level 4 1,506 edu level 5 3,443 edu level 6 1,063
a. Dependent Variable: CASP: quality of life and well-being index
VIF quantifies the severity of multicollinearity in regression analysis; higher VIF’s have higher multicollinearity. As shown above, VIF- the abbreviation for variance inflation
factor- values range from 4,32 for the third educational level to 1,06 for the sixth level of education, maximum and minimum value respectively. Education levels 1,2,3 and 5 all have VIF>3, whereas education levels 4 and 6 display VIF levels lower than 2. These multicollinearity diagnostics for education level show that there is in fact some
multicollinearity between education levels. Levels 1,2,3 and 5 all have VIF’s larger than 3, which is generally thought to propose a problem. However, as noted by Allison (2012) if dummy variables are used and if the proportion of cases in the reference category is small, the indicator variables will necessarily have high VIFs. This entails that this research can safely ignore multicollinearity because it uses dummy variables for both education and career choice.
Table 3: Multicollinearity report of job industries
Job industry VIF
agriculture, hunting, forestry, fishing 1,126
mining and quarrying 1,017
manufacturing 1,190
electricity, gas and water supply 1,034
construction 1,099
wholesale and retail trade 1,132 hotels and restaurants 1,028 transport, storage and communication 1,076 financial intermediation 1,041 real estate, renting and business activities 1,015 public administration and defence 1,122
education 1,119
health and social work 1,127 a. Dependent Variable: CASP: quality of life and well-being index
VIF levels for job industries show a small range from 1,02 till 1,19. Career choice VIF’s are all lower than 1,20 so multicollinearity is not an issue here to begin with. In addition, because this research uses a very large sample (N>20000), multicollinearity is less of an issue because of statistical power (Stata example, 2014).
5.2 Summary statistics
First of all, some comments on the statistics, such as percentages and means of the variables in this research are made. The majority of the respondents is female (55,20%), with males filling out the other 44,80%. The valid N (N=number of respondents) of gender is 26347. The average age of the respondents is 74 years old, with a standard deviation (SD) of 10,20. The high average age in the data is a result of SHARE’s policy that only individual’s aged 50 years and up are eligible to participate in the study.
Table 4: Gender respondent frequencies
Frequency Percent Valid Percent* Cumulative Percent
Valid
Male 11802 44,8 44,8 44,8
Female 14545 55,2 55,2 100,0
Total 26347 100,0 100,0 *Valid percent excludes missing cases
The median of education is level 3, with a SD of 1,43. Only people who completed at least primary education level are included. Out of the valid cases, frequency ranges from 0,50% of the respondents having level 6 education and 30,6% of respondents having level 3 education. The median years of education is 11 with a SD of 4,35. Maximum and minimum years of education were 25 and no years respectively. 21873 respondents filled in their ISCED-97 scale information. 4474 values are missing and excluded from the results.
Table 5: Education of respondent in ISCED-97 Coding
Frequency Percent Valid Percent* Cumulative Percent Valid isced-97 code 1 6112 23,2 27,9 27,9 isced-97 code 2 4033 15,3 18,4 46,4 isced-97 code 3 6685 25,4 30,6 76,9 isced-97 code 4 716 2,7 3,3 80,2 isced-97 code 5 4228 16,0 19,3 99,5 isced-97 code 6 99 ,4 ,5 100,0 17
Total 21873 83,0 100,0
Missing System 4474 17,0
Total 26347 100,0
*Valid percent excludes missing cases
Job industry, which ranges from 0,90% working in “real estate, renting and business activities” to 18,10% working in manufacturing. 10,60% of respondents falls in the category “other”. “Agriculture, hunting, forestry and fishing”, “wholesale and retail trade”, “public administration and defence”, “education” and “health and social work” all have more than 10 percent of the respondents working in them.
Table 6: Job industry frequencies and respective job industry percentages of total.
Frequency Percent Valid Percent Cumulative Percent
Valid
agriculture, hunting,
forestry, fishing 1889 7,2 11,1 11,1 mining and quarrying 245 ,9 1,4 12,5 manufacturing 3079 11,7 18,0 30,5 electricity, gas and water
supply 464 1,8 2,7 33,3
construction 1496 5,7 8,8 42,0
wholesale and retail trade 2064 7,8 12,1 54,1 hotels and restaurants 375 1,4 2,2 56,3 transport, storage and
communication 1092 4,1 6,4 62,7
financial intermediation 587 2,2 3,4 66,1 real estate, renting and
business activities 218 ,8 1,3 67,4 public administration and
defence 1869 7,1 10,9 78,4
education 1762 6,7 10,3 88,7
health and social work 1930 7,3 11,3 100,0
Total 17070 64,8 100,0
Missing System 9277 35,2
Total 26347 100,0
On a scale from 12 to 48, the mean of happiness is 37,12, with the median being 38. For every point on the scale there were respondents representing them. CASP-12 information is available for 22126 respondents. Table 7 shows a quick summary of all the used variables, including means, medians and standard deviations for easy reference. For further reference, the tables with the frequencies are appended. The appendix will also contain bar charts that visualize the distributions of the following variables: gender, job industry, happiness and education.
Table 7: Statistics of used variables gender respondent
job industry education of respondent in ISCED-97 Coding years of education CASP: quality of life and well-being index N Valid 26347 17070 21873 22929 22126 Missing 0 9277 4474 3418 4221 Mean 6,97 2,69 10,5200 37,12 Median 1,00 6,00 3,00 11,0000 38,00 Std. Deviation 4,35372 6,194 Skewness ,100 -,496 Std. Error of Skewness ,016 ,016 Kurtosis -,051 -,166 Std. Error of Kurtosis ,032 ,033 Minimum 0 1 1 ,00 12 Maximum 1 13 6 25,00 48 19
Table 8: Mean of happiness for the different education levels
Education level Mean of happiness
Edu level 1 34,82 Edu level 2 37,26 Edu level 3 38,15 Edu level 4 38,61 Edu level 5 39,51 Edu level 6 38,29
The mean of happiness increases from education level 1 through 5, with minimum and maximum values being 34,82 and 39,51 respectively. After level 5, the mean of happiness takes a slight drop back to 38,29 on the CASP-12 scale. So there is a clear increase apparent on educational level till we reach level 6. The average of happiness tops out at 39,51 for education level 5, which is the first stage of tertiary education. With N=6685, upper-secondary education (level 3) is the median of education, followed by the least happy educated individuals, level 1 education with N=6112. Figure 3 visualizes this relationship between educational attainment and happiness.
Figure 4: Line chart of the educational attainment and happiness relationship
Table 9: Mean of happiness for the different job industries
Job industry Mean of happiness agriculture, hunting, forestry, fishing 35.03117
mining and quarrying 36.69000
manufacturing 37.64671
electricity, gas and water supply 37.68238
construction 37.46575
wholesale and retail trade 37.94954 hotels and restaurants 36.45231 transport, storage and communication 37.71872 financial intermediation 39.11317 real estate, renting and business activities 39.86364 public administration and defence 38.22589
education 38.96732
health and social work 39.21731
Happiness means for the different job industries range from values between 35 and 40. Average happiness is highest for the real estate, renting and business activities industry, lowest happiness levels are measured for the agriculture, hunting, forestry and fishing sector.
5.3 ANOVA
The analysis of variance between education and quality of life is shown to be highly significant (α=0.0001). This proves there are indeed differences between groups for the independent variable education and the dependent variable quality of life (CASP).
Table 10: ANOVA table of educational attainment and CASP: quality of life and well-being index
Sum of Squares df Mean Square F Sig. Between Groups 61119,722 5 12223,944 360,117 ,000
Within Groups 708045,006 20859 33,944 Total 769164,728 20864
The Analysis of variance between job industry and quality of life (CASP) shown below is again proven to be significant with α=0.0001. This proves, again, that there are differences between groups.
Table 11: ANOVA of job industry and CASP: quality of life and well-being index
Sum of Squares df Mean Square F Sig. Between Groups 20907,001 12 1742,250 52,379 ,000
Within Groups 483733,401 14543 33,262 Total 504640,403 14555
5.4 Regression analyses
5.4.1 Regression models
Table 12: ANOVA table of the linear regression model with education level (IV) as independent variable and CASP (DV) as the dependent variable
Model Sum of Squares df Mean Square F Sig.
Regression 79865,753 6 13310,959 382,898 ,000b 22
Residual 768939,570 22119 34,764 Total 848805,323 22125
a. Dependent Variable: CASP: quality of life and well-being index
b. Predictors: (Constant), edu level 6, edu level 4, edu level 2, edu level 5, edu level 1, edu level 3
The regression model for education level and CASP is found to be significant on the α=0.001 level.
Table 13: ANOVA table of the linear regression model with job industry (IV) and CASP (DV)
Model Sum of Squares df Mean Square
F Sig.
Regression 1370682,193 13 105437,092 2757,126 ,000b Residual 56932362,674 1488752 38,242
Total 58303044,867 1488765
a. Dependent Variable: CASP: quality of life and well-being index
b. Predictors: (Constant), health and social work, real estate, renting and business activities, mining and quarrying, hotels and restaurants, electricity, gas and water supply, financial intermediation, transport, storage and communication, education, construction, public administration and defence, wholesale and retail trade, agriculture, hunting, forestry, fishing, manufacturing
The regression model for job industry and quality of life is also found to be significant (α=0.001).
5.4.2 Regression analyses
Table 14: Coefficients of dummy coded education levels and CASP (DV)
Model Unstandardized Coefficients Standardized coefficients t Sig. Collinearity Statistics
B Std. Error Beta Toleranc
e VIF (Constant) 33,378 ,166 201,029 ,000 edu level 1 1,442 ,183 ,102 7,863 ,000 ,243 4,116 edu level 2 3,885 ,191 ,238 20,319 ,000 ,298 3,355 edu level 3 4,771 ,182 ,349 26,268 ,000 ,231 4,321 edu level 4 5,235 ,278 ,148 18,836 ,000 ,664 1,506 edu level 5 6,136 ,190 ,383 32,273 ,000 ,290 3,443 edu level 6 4,916 ,661 ,049 7,440 ,000 ,940 1,063 a. Dependent Variable: CASP: quality of life and well-being index
Beta’s of the different levels of education and the dependent variable happiness bottom out at ,049 for education level 6 and have a maximum of ,383 for education level 5. The second highest beta is education level 3 and the second lowest beta is for education level 1. All values are significant at α=0.001. The standardized coefficient betas are the estimates on the
independent variable that are standardized to have a variance of 1. This simulates how many standard deviations the dependent variable changes when the predictor variable changes with 1 standard deviation. For example, for every standard deviation change in education level 1, the predictor variable happiness changers with ,102 standard deviations. The main benefit over the unstandardized method is the added simplicity it brings to interpretation. Tolerance value originates from dividing 1 by the before mentioned VIF value. This regression with dummy coded education levels provides the final prove for a causal relationship between education and happiness. The outcomes of this statistical measure further support the previous findings that happiness increases with an increase in educational attainment. Level 5 of
education is found to explain 38,30% of the variance in happiness. With 4,90% of explained variance, the highest level of education has the lowest incremental effect on happiness.
The first hypothesis in this study stated that: Level of Education is positively
correlated to level of happiness. The outcomes of the analysis fully support this hypothesis, education level does positively correlate with happiness, and the regression analysis shows a relationship between these variables.
Sub-hypothesis 1a: Happiness increases for every incremental level of education. This hypothesis is not supported by the data, for the first five levels of education it is found to be true, but because the sixth level of education takes a small dive in average happiness this hypothesis must therefore be rejected. Figure 4 shows this relationship.
5.4.3 Comparing the education and happiness relationship to existing literature The findings complement extensive past research that also found education to positively correlate with happiness. The effect in this study is higher than most of the other studies in this field, Fowler, Christakis, Steptoe & Roux (2009, p.2) found that educated people are only slightly happier. Especially for level 5 educational attainment, which has a high explained variance of happiness close to 40% is an interesting result. Still open for discussion is whether the results of a better education, such as higher income and job opportunities, improve the happiness rating and not the education itself. This is often suggested in past research, Boehm and Lyubomirsky (2008) write: “happiness often precedes measures of success”. Research by del Mar Salinas-Jiménez Artés and Salinas-Jiménez (2011) contradicts this suggestion, it states that education has a significant effect on life satisfaction, independent of its effect on income. Pischke’s (2011) estimates suggest that most of the association of income and well-being is causal, leaning towards the idea that the benefits resulting from education improve happiness, and not education itself. More research could clarify some of the underlying reasons for this relationship.
5.4.4 Regression of job industries and quality of life
Table 15: Coefficients of dummy coded job industries and CASP (DV) Model Unstandardized Coefficients Standardized Coefficients t Sig. Collinearity Statistics
B Std. Error Beta Toleran
ce VIF (Constant) 35,830 ,070 515,49 2 ,000 agriculture, hunting, forestry, fishing -,799 ,165 -,034 -4,846 ,000 ,888 1,126 25
mining and quarrying ,860 ,433 ,013 1,985 ,047 ,983 1,017 manufacturing 1,817 ,136 ,096 13,349 ,000 ,841 1,190 electricity, gas and water
supply 1,852 ,309 ,040 5,991 ,000 ,967 1,034 construction 1,636 ,185 ,061 8,831 ,000 ,910 1,099 wholesale and retail trade 2,119 ,161 ,092 13,133 ,000 ,883 1,132 hotels and restaurants ,622 ,343 ,012 1,816 ,069 ,973 1,028 transport, storage and
communication 1,889 ,210 ,061 9,009 ,000 ,929 1,076 financial intermediation 3,283 ,283 ,078 11,601 ,000 ,961 1,041
real estate, renting and
business activities 4,034 ,461 ,058 8,747 ,000 ,985 1,015 public administration and
defence 2,396 ,167 ,099 14,308 ,000 ,891 1,122 education 3,137 ,170 ,129 18,507 ,000 ,894 1,119 health and social work 3,387 ,164 ,144 20,626 ,000 ,887 1,127 a. Dependent Variable: CASP: quality of life and well-being index
The majority of the beta’s of job industry on the dependent variable happiness are smaller than the beta’s of educational attainment on happiness. It includes one negative value of -,034 for “agriculture, hunting, forestry, fishing”. The positive values range from ,012 for hotels and restaurants till a beta of ,144 for health and social work. Education closely follows health and social work with a beta of ,129. All values are significant at α=0.001, except for mining and quarrying, which is significant at the 5% level and hotels and restaurants which is only significant for the α=0.1 level. Before, we looked at the means of happiness for the different job industries and they seemed to support differences between groups. Regression analyses show low happiness variances explained by career choice. One negative value of -,034 for agriculture , hunting, forestry and fishing is produced. The highest explained variances are contributed to the job industries “education” and “health and social work”, with respective betas of ,129 and ,144.
Career choices affect happiness, was the second hypothesis formulated in this study. The findings support this hypothesis, although only relatively small effects are noticed. It is noted
that apart from one, all career choices improve happiness ratings. Because “other” industries were left out, it seems the jobs in this industry negatively affect happiness, but because they are beyond the scope of this research it has to be left to future research to look into this.
5.4.5 Comparing the career choice and happiness relationship to existing literature The findings neither support nor reject the outcomes of past literature. Career choices have been found to have effects on happiness before, but they were directed more on an individual level and not on the industry level. Henderson (2000) found that people who follow their instincts achieve career happiness, but this does not necessarily relate to choosing a job industry. Consequently, more research has to be done to arrive to any conclusions regarding the effect of career choice on happiness, and possible causalities influencing this effect.
5.5 Interaction effects
Table 16 lists all the interaction terms, their estimates, significance and the corresponding mean of happiness.
Table 16: Interaction coefficients: (2 not defined because of singularities)
Estimate Std. error t-value Pr(>|t|) Mean of happiness (Intercept) 33.72626 0.19562 172.410 < 2e-16 *** 33.73 isced_r2 3.24330 0.41617 7.793 6.99e-15 *** 36.97 isced_r3 4.23127 0.39760 10.642 < 2e-16 *** 37.96 isced_r4 3.88912 1.10990 3.504 0.000460 *** 37.62 isced_r5 5.00843 0.59577 8.407 < 2e-16 *** 38.73 isced_r6 1.27374 5.57423 0.229 0.819257 35.00 industry2 1.79374 0.81175 2.210 0.027142 * 35.20 industry3 2.42292 0.29001 8.355 < 2e-16 *** 36.15 industry4 1.73949 0.68072 2.555 0.010619 * 35.47 industry5 2.01805 0.35557 5.676 1.41e-08 *** 35.74 industry6 2.47374 0.34788 7.111 1.21e-12 *** 36.20 industry7 1.96681 0.58782 3.346 0.000822 *** 35.69 27
industry8 2.76222 0.42577 6.488 9.02e-11 *** 36.49 industry9 4.08855 1.08980 3.752 0.000176 *** 37.81 industry10 5.14874 1.40637 3.661 0.000252 *** 38.88 industry11 1.89002 0.46765 4.042 5.34e-05 *** 35.62 industry12 1.45429 0.68505 2.123 0.033779 * 35.18 industry13 3.87691 0.44996 8.616 < 2e-16 *** 37.60 isced_r2:industry2 -3.48125 1.26080 -2.761 0.005767 ** 35.28 isced_r3:industry2 -2.20715 1.11137 -1.986 0.047056 * 37.54 isced_r4:industry2 -1.40912 2.65044 -0.532 0.594973 38.00 isced_r5:industry2 -3.22408 1.52476 -2.114 0.034492 * 37.30 isced_r6:industry2 5.20626 6.87092 0.758 0.448629 42.00 isced_r2:industry3 -2.24856 0.52635 -4.272 1.95e-05 *** 36.15 isced_r3:industry3 -1.93684 0.48709 -3.976 7.03e-05 *** 37.14 isced_r4:industry3 -0.44795 1.28515 -0.349 0.727428 38.44 isced_r5:industry3 -1.32380 0.70090 -1.889 0.058950 . 39.59 isced_r6:industry3 -1.22292 6.10938 -0.200 0.841350 36.20 isced_r2:industry4 -2.65500 1.00881 -2.632 0.008502 ** 35.47 isced_r3:industry4 -1.17233 0.88021 -1.332 0.182925 36.05 isced_r4:industry4 -2.75487 1.79128 -1.538 0.124087 38.52 isced_r5:industry4 0.05707 1.12470 0.051 0.959534 36.60 isced_r6:industry4 NA NA NA NA isced_r2:industry5 -1.99752 0.64420 -3.101 0.001934 ** 36.99 isced_r3:industry5 -1.26022 0.56958 -2.213 0.026946 * 38.72 isced_r4:industry5 -0.66122 1.47719 -0.448 0.654435 38.97 isced_r5:industry5 -1.34098 0.79098 -1.695 0.090033 . 39.41 isced_r6:industry5 -8.01805 7.88631 -1.017 0.309311 29.00 isced_r2:industry6 -1.34710 0.57838 -2.329 0.019868 * 38.10 28
isced_r3:industry6 -1.83654 0.53762 -3.416 0.000637 *** 38.60 isced_r4:industry6 -1.28061 1.40532 -0.911 0.362174 38.81 isced_r5:industry6 -1.42536 0.77582 -1.837 0.066196 . 39.78 isced_r6:industry6 3.52626 6.44200 0.547 0.584122 41.00 isced_r2:industry7 -2.37166 0.91954 -2.579 0.009914 ** 36.56 isced_r3:industry7 -2.09507 0.91859 -2.281 0.022579 * 37.83 isced_r4:industry7 -1.45719 2.32774 -0.626 0.531317 38.13 isced_r5:industry7 -4.22002 1.34596 -3.135 0.001720 ** 36.49 isced_r6:industry7 4.03319 7.90019 0.511 0.609696 41.00 isced_r2:industry8 -2.68203 0.68601 -3.910 9.29e-05 *** 37.05 isced_r3:industry8 -2.28482 0.63215 -3.614 0.000302 *** 38.44 isced_r4:industry8 -2.62760 1.57583 -1.667 0.095449 . 37.75 isced_r5:industry8 -2.11631 0.85413 -2.478 0.013234 * 39.38 isced_r6:industry8 NA NA NA NA isced_r2:industry9 -2.19448 1.33895 -1.639 0.101246 38.86 isced_r3:industry9 -3.20876 1.20664 -2.659 0.007840 ** 38.84 isced_r4:industry9 -0.51644 1.83059 -0.282 0.777860 41.19 isced_r5:industry9 -3.47073 1.31438 -2.641 0.008286 ** 39.35 isced_r6:industry9 -6.08855 6.90929 -0.881 0.378218 33.00 isced_r2:industry10 -2.61830 1.87708 -1.395 0.163074 39.50 isced_r3:industry10 -3.79377 1.60705 -2.361 0.018254 * 39.31 isced_r4:industry10 -1.36412 2.50497 -0.545 0.586060 41.40 isced_r5:industry10 -3.24192 1.69708 -1.910 0.056116 . 40.64 isced_r6:industry10 3.05126 6.26245 0.487 0.626102 43.20 isced_r2:industry11 -0.87974 0.69192 -1.271 0.203593 37.97 isced_r3:industry11 -1.78862 0.62511 -2.861 0.004225 ** 38.06 isced_r4:industry11 -1.05085 1.37204 -0.766 0.443747 38.45 29
isced_r5:industry11 -0.96392 0.77700 -1.241 0.214786 39.66 isced_r6:industry11 -0.39002 5.86138 -0.067 0.946949 36.50 isced_r2:industry12 -0.26655 0.97617 -0.273 0.784811 38.16 isced_r3:industry12 -1.68132 0.83616 -2.011 0.044369 * 37.73 isced_r4:industry12 0.63699 1.44107 0.442 0.658477 39.71 isced_r5:industry12 -0.54973 0.90435 -0.608 0.543287 39.64 isced_r6:industry12 2.25159 5.77309 0.390 0.696531 38.71 isced_r2:industry13 -1.95600 0.69654 -2.808 0.004990 ** 38.89 isced_r3:industry13 -2.40705 0.62183 -3.871 0.000109 *** 39.43 isced_r4:industry13 -2.97980 1.33566 -2.231 0.025700 * 38.51 isced_r5:industry13 -2.62438 0.75413 -3.480 0.000503 *** 39.99 isced_r6:industry13 -0.43691 5.69891 -0.077 0.938891 38.44 Significance codes: ***α=0.001,** α=0.01, * α=0.05, .α =0.1
The interaction terms produce mixed results, 33 of the interactions are non-significant
(although 5 of them are significant on the p<0.1 scale), the other 42 are flagged as significant. The significant estimate values lie between -4.22 and 5.01. There are 2 combinations that are non-existent, they are marked NA. These are the isced_r6*indsytry4 combination and the isced_r6*indsytry8 combination. The (significant) means of happiness hover between 35.18 for isced_r1*industry12 and 39.99 for isced_r5*industry11. Current streams of research suggest that career choices affect happiness, although they don’t necessarily use job industry as a measure of career choice. Added, no literature seems to exist on the
education*careerchoice interaction. Even though the interaction term produced insignificant values for the first time in this study, the majority still had significance. The interaction outcomes seem to further build on the proof that higher education levels lead to higher happiness. Interaction terms with level 5 education mostly produced the highest happiness results, followed by the fourth and sixth levels of education respectively. The lowest
educational attainment, level 1, corresponded to the lowest levels of happiness, ranging from 35.20 to 38.88. Hereby, the third hypothesis is supported, there is indeed an interaction effect between educational attainment and career choices on happiness, but not across all
interactions (66% of the values are significant) and the explained variances for the interaction
terms are also small. The implications of this finding among the other results are outlined below in the theoretical and practical implications.
5.6 Theoretical implications
With most hypotheses supported, this research has some theoretical implications. Again, education is found to increase happiness levels in individuals, adding to the already extensive line of research done on the subject (Noval, Garvi, year unknown; Powdthavee, Lekfuangfu & Wooden 2013; Hartog and Oosterbeek, 1997). However, this study finds education level to vary substantially across the different educational attainment levels and in addition it also finds a higher correlate than most other research. Thus suggesting that education is an even more important factor in explaining happiness than past research suggests. Surprisingly, the highest level of education is found to decrease happiness somewhat. Level 6 education corresponds to degrees such as doctorates.
Career choice is found to significantly impact happiness. They explain small variances between -3,4% and 14,4%. Implications are that career choice is indeed an important factor in quality of life. Past literature seemed to fit with this outcome. Some job industries
significantly undercut the mean of happiness, where some industries structurally outperform quality of life. To my knowledge, linking job industry to happiness is new in research and it provides a new theory on factors that influence quality of life.
Again seemingly new to current theory is the interaction of educational attainment and happiness. Mixed results, positively and negatively affecting happiness and not always
significant, it still deems itself worthy to research, adding to the separate variables that influence happiness.
5.7 Practical implications
Next to theoretical implications, this research has some practical implications as well. Debatably the most important is the implications it has for students and their life choices. A fairly substantial part of happiness is explained by education (26,6%), further adding to existing literature, proving that students benefit from higher education levels. If happiness is an important factor in life, and for 69% of people it is the single most important thing in life (Diener, 2000), then they might feel enlightened by this research. Suggesting that maybe too much educational attainment reverses the positive happiness effect is an important sidenote to keep in mind. The lion’s share of students might already know these facts, but maybe they are unaware of the forthcoming changes career choice has on happiness. Findings suggest
positive and negative linear relationships between career choice and happiness, and it serves students to look what there dreamed career paths imply for average happiness levels. With
this research, students can look up what combinations of education and career choice lead to higher happiness levels and use this info as a decision making basis. Nevertheless, this research should only be used as an addition to other information on making the right life choices. Instinct, character and culture should not be neglected in these choices.
Happiness is not only important to individuals themselves, but also to the enterprises they work in. Happy employees consistently outperform their unhappy colleagues in
productivity and helpful acts (Boehm & Lyubomirsky, 2008). It is therefore crucial for businesses to keep their employees content. The outcomes of this research can be used by managers around the world to give the right jobs to the right people. Looking at the education of current employees and job applicants, job openings can be given to the people that
potentially get the most out of the job in terms of happiness. This can create a sustained competitive advantage to the company. This study can also partly explain why people are unhappy. They might have chosen a career choice that influenced their quality of life in a negative manner, starting another study or choosing another job could positively impact their happiness.
6. Conclusion
This study focused on the evaluation of education and career choice and their influences on happiness to try and clarify some characteristics of the labour market, which was the primary focus of this research. More specifically, focused laid on understanding the moderating effect of career choice on the already well-established line of research on education and happiness. Open data from SHARE was used to evaluate these propositions. Using statistical means the various hypotheses were tested. Results were compared and matched to the literature review to unveil any contradictions and similarities.
Congruent to existing literature (Noval, Garvi, year unknown; Powdthavee,
Lekfuangfu & Wooden 2013), the findings on the education happiness relationship imply that there is a positive correlation between educational attainment and happiness. Small
relationships between job industries (also referred to as career choices) and happiness were found, both negative and positive. The final findings, on the moderating effect of career choice on the education happiness relationship, suggest that career choice is indeed a moderating factor. As a result, all hypotheses are supported by the findings.
“Individuals pursue a higher education to acquire adequate knowledge for the dream job they always wanted.” This statement, the opening sentence of this research looks to be confirmed by the findings. Students can use these findings in the future to aid them in
choosing their studies and careers in the hope of one day working their dream job and achieving what 69% of people find most important in life: happiness (Diener, 2000).
7. Recommendations for future work
This final part of the study will firstly list limitations specific to this study so future research has a clear understanding of what is lacking this research and where there is room for improvement. Finally, suggestions for future research are outlined to incentivize researchers to find new causal relationships and encourage in depth research using more demographics, especially on a geographical level. Using more job specific data, e.g. individual jobs could yield better results. A statement meant for future researchers concludes the recommendations.
7.1 Suggestions for future research
Future research could use and is encouraged to use the wide range of variables concluded in the SHARE databases, this research has 85 variables from a combination of 2 databases. Correcting for the before mentioned factors influencing happiness (Income, health, marital status) is one of the options that could yield better results. Adding variables to explain more of the happiness variance and improving the model is another option. Gender is perceived to have a small effect on happiness and could be added to further research in this field.
Separating the job industries in individual jobs could improve results by explaining more of the happiness variance. Modelling with job type, such as manual labour or voluntary work instead of job industry could lead to different outcomes.
Doing the same analyses but in other regions possibly outputs different results, especially developing countries, where data is more restrictive in terms of size and
accessibility. Future research is suggested to look into the different demographics. One of few conflicting results was the dip in happiness of the highest level of education (level 6)
compared to the preceding level 5 educational attainment. Future research could look at factors underlying this change. Also, the high explained variance explained by the fifth level of education on happiness (38%) is an interesting result. Research could dig to find reasons for this high explained variance.
Future research is invited to elaborate on and improve the results of this study. How does career choice influence happiness? Future models that explain more on the relationships included in this study are a welcome addition because a lot is still unclear on labour market characteristics, education and the implications they have on quality of life. This research tried
to bridge the gap that exists in happiness studies, namely the lack of research on the relationship between career choice and happiness. Future research hopefully jumps on the bandwagon to add to this research and improve knowledge on the labour market and quality of life studies.
