Effect of Age and Education on Lifetime Employment
Niels van der Veer (10000091)
University of Amsterdam
Bachelor's Thesis
BSc Economics and Business Administration Thesis Seminar Business Administration
Supervisors: Vladimer Kobayashi, Stefan Mol and Gabor Kismihok June 29, 2015
Statement of Originality
This document is written by Student Niels van der Veer who declares to take full
responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is
Table of Contents
Effect of Age and Education on Employment... 1
Background of the Study... 1
Goals and Objectives... 1
Scope and Limitations... 2
Impact and Contribution... 2
Literature review... 2 Conceptual framework... 7 Model... 7 Hypotheses... 8 Methodology... 8 Database exploration... 8 Eurostat... 9
The World Bank Group... 10
Quandl ... 10 LABORSTA... 10 ILOSTAT... 10 Variables... 11 Age... 11 Employment... 11 Educational attainment... 11 Analyses... 13
Results and discussion... 15
Correlations... 19
Kruskal-Wallis H test... 21
Friedman test... 23
Moderator analysis... 24
Conclusion... 27
Recommendations for future research... 29
Abstract
Previous studies haveresearched the relationship between age, education and
employment. However, to my knowledge, this relationship has not been fully explored in the Netherlands. Research done by Ashenfelter and Ham (1979) and Biagi and Lucifaro (2008) found positive effects of age and education on employment. Firstly, I hypothesized that if a person gets older his/her employment chance increases up to a certain age. Once this age is reached employment chance tends to decrease. Secondly, I hypothesized that a person who attained a high educational attainment level has a higher employment probability than a person who attained a lower educational attainment level. Thirdly, I proposed that the
education level has an influence on the relationship between age and employment. The study tried to find support for these hypotheses by using existing employment rate data from EUROSTAT (N=3837). In this study age was clustered in 11 groups with a five year range per group. Education was divided in three groups using the ISCED 2011 scale levels. Age was found to have a positive effect on employment, but not for persons between 60 and 69. This is probably because those people are not looking for jobs anymore. Also the level of education positively influenced the employment probabilities of a person. Lastly, no interaction between age and educational attainment level was found. This indicates that the positive effect of age on employment is not influenced by the education level a person attained.
1 Effect of Age and Education on Lifetime Employment
The government recently transformed the study financing system into a loan system and therefore the average study costs rose to thousands of euros (Verhoef, 2014). This forces adolescents to make a decision between studying at high costs or to quit education early and start working. A reason for people to leave education is that a person is not able to cope with
the high study costs and instead starts to earn money by getting a job in an earlier stage of his/her life that he/she would prefer. But is this the right decision?
Earlier studies suggest that unemployment rates are inversely related to the educational attainment levels of workers (Ashenfelter & Ham, 1979). This would mean that highly educated people are on average more employable throughout their lives than low or medium educated people, even though their education lasts longer than those of the lower educated people. Because in the coming years the retirement age is slowly going up in the Netherlands, from the age of 65 to 67, the Dutch government forces people to work longer (Rijksoverheid, 2014).
In this study I want to investigate if age positively influences employment and if this positive influence varies across different educational attainment levels by trying to find what the effect of age is on employment in the Netherlands and if the educational attainment of workers has an influence on this effect.
Goals and Objectives
To begin this study open data sources need to be explored. Multiple open databases can be found online that contain datasets about education and the labour market. When useful data sources are found I will research if the age of a person has a positive influence on his/her chances of being employable in the Netherlands. If there is statistical evidence found, then the variation of this potential positive influence across different education levels is investigated
2 by doing post hoc analysis. Therefore, the goal in this study is to find the answer to the following research question: 'Does age positively influence employment in the Netherlands
and does this influence differ across different educational attainment levels?
In this study I wanted to find if there is a positive relationship between age and employment and if there was, if the educational attainment level of a person moderated this relationship.
Scope
The scope of this study was the use of data from the Netherlands. The used dataset contained data about people between the age of 15 and 69. Moreover, all different educational levels attained were used. Because the focus was on the working population in the Netherlands, persons from other countries were not used.
Impact and Contributions
This study will contribute to research on the relation between employment and age by specifically searching for the effect of educational attainment on that relationship. If statistical evidence is found that shows that people with higher educational attainment are indeed more employable throughout their lives, then students could consider making extra study costs. If no evidence for the age and employment relationship is found, then it means that students probably should not consider higher educational attainment just for the reason to have the certainty of having a job at a later stage in their lives, because attaining a higher education level than secondary education makes no difference in being employable.
Literature
Unemployment rates are rising across the world and the Netherlands have been abiding this trend (Centraal Bureau Statistiek [CBS], 2014). In the past 5 years the unemployment of Dutch people in the working population has almost doubled. In 2009 the
3 percentage of unemployed Dutch citizens was 4.8% and in the first quarter of 2014 it has risen to 9.0% (CBS, 2014). The unemployment issue causes worries for the Dutch government as well as for the Dutch working population. Statistics show that the unemployment in the Netherlands has grown overall, but do the rates differ significantly among different age groups? The facts show that they do for the European Union. According to numbers published by the European Commission in 2012 the youth unemployment in Europe was more than twice as high than the unemployment amongst adults 23.3% and 9.3% respectively in the fourth quarter of 2012 (European Commission, 2014). One explanation to account for these figures is that the majority of people with the age range 15 to 25 with a job in Europe, and especially in the Netherlands, are working part-time. In 2013 78% of the youth were working part-time against 50% for the total working population in the Netherlands (Deeltijdwerk in Europa neemt, Nederland koploper; 2014). Since part-time and temporary contracts are less stable than full-time contracts, this is a cause for higher unemployment amongst the young working people.
A general finding appears to be that education reduces the incidence and possibly the duration of such unemployment for an individual (Ashenfelter and Ham, 1979; Devine and Kiefer, 1991; Kiefer, 1985; Nickell, 1979). Education expands employment opportunities since educated workers are productive in all jobs whereas less educated workers are productive in only some jobs (McKenna, 1996). An important theory is the one of human capital. Human capital is seen as any stock of knowledge or characteristics the employee has that increases his/her productivity. It is a marketable set of skills in which employees invest to differentiate themselves on the labour market. One of these skills is knowledge attained from education. Olaniyan and Okemakinde (2008) state that education is an economic good, because it is not easily obtainable and thus needs to be apportioned. Furthermore, education is
4 a form of input into the production of other goods and services. Human capital theorists even argue that an educated population is a productive population. Employees improve their productivity and efficiency by increasing their knowledge on economical productivity, which requires an investment done by the employee, i.e. the costs and time spent on education. A higher educational attainment level therefore leads to higher employability. Employees who attain higher education levels are seen as greater forms of human capital and therefore more valued by employers. Because of the increased economical productivity that comes with education, educated people are preferred over uneducated people in jobs that also the uneducated are qualified for. Educated people often choose to have such a job while searching for a job that meets their qualifications. This simultaneously causes higher unemployment for the lower educated and lower unemployment and shorter unemployment durations for the higher educated (McKenna, 1996). Numbers from the European Commission confirm this by showing that early leavers from education and training are indeed a high-risk group. The unemployment rate for this group was 55.5% in 2013, although 70% of the persons in this group wanted to work (European Commission, 2014). This argument does not only count for the people between the age of 15 and 25, but it is a cause of the employment differences between high and less educated people throughout a lifetime. A reason for people to leave education early can be the high study costs, the lack of intelligence or a bad fit with the education system. Human capital theory not only accounts for education. All sets of skills that a person acquires during his lifetime can be seen as human capital. Often skills are learned by experience and experience comes with the time you spend on something. Therefore it is also expected that age positively relates to employability. The older a person is, the more work experience the person has and the more valuable he/she is to the employer. But it is also expected that people start to lose some skills from a certain age. People lose productivity and
5 efficiency, because they become physically less flexible or lose some of their cognitive capabilities. This means that older people decrease in value of human capital from a certain point of age.
A study by Biagi and Lucifora (2008) was performed on the unemployment in ten European countries (excluding the Netherlands). They researched the differences in the unemployment rates of people between 1991 and 2002 by separating the working population in two age groups and in two education groups. In the first part they found that in nine of the ten European countries the youth unemployment (age 15-24) was more than twice as high than for the adults group (age 25-54). For Germany the unemployment rates were equal. In the second part they investigated if the group with people who attained secondary education and higher have lower unemployment rates than the group who attained no or only primary education. The conclusion was that in Spain and Sweden the unemployment rates of both groups were equal. In France, Greece, Italy and Portugal the unemployment was higher amongst higher educated people and in Finland, Germany, Norway and the United Kingdom it was higher amongst the lower educated. There is no real evidence in current literature that education is a predictor for better employment rates, but it supports the earlier studies which found that there were lower unemployment rates for adults than for adolescents.
The reason for the higher employment of older people is probably the before mentioned reason that they have more work experience and skills and are therefore preferred over younger people. Since older people are already preferred it is expected that if older people attain higher education levels their employability increases even more than that of young people. In other words, younger people attaining higher education will not be preferred over the older, more experienced people. Therefore the effect of age on employment will increase if the education level increases.
6 Earlier studies were found on unemployment differences between adolescents and adults and on the relation between employment and education. Those studies were performed in a variety of contexts, but to our knowledge it has not yet been performed in the Netherlands. Therefore in this paper the subject is researched in the Netherlands to find if the results are in line with the results from previous studies.
A two-way ANOVA compares the mean differences between groups that are divided by two independent variables. It helps to understand if there is an interaction between the two independent variables on the dependent variable. The interaction term shows if the effect of one independent variable on the dependent variable is the same for all values of the other independent variable. To perform a two-way ANOVA some assumptions need to be met. The first assumption is that the dependent variable should be measured at the continuous level. Since employment rates are continuous (from 0,00% to 100,00%), this assumption is met. The second is that the independent variables should consist of at least two categorical, independent groups. This assumption is met as well, because employment is compared between different educational attainment levels and different age groups. Thirdly, there is independence of observations. This is also the case, since participants cannot be in two age groups or at two educational attainment levels simultaneously. The fourth assumption is that there are no significant outliers. Since those are removed, this assumption is met as well. The fifth assumption is that the distribution of the dependent variable should be approximately normally distributed for each combination of groups of the two independent variables. This assumption is violated, because for every condition the employment is not normally distributed. A Shapiro-Wilk test showed a p < 0.001 for each age group which indicates that employment is not normally distributed. The last assumption is the one of homogeneity of variances for each combination of the groups of the two independent variables (Laerd
7 Statistics, 2015). This assumption is met as well, since there is homogeneity for each combination of the groups of the two independent variables.
Because not all assumptions are met, an alternative to the two-way ANOVA needs to be chosen. An alternative that does not have the assumption of normally distributed data is the Friedman's test. This nonparametric test is designed for exactly one measurement for each possible combination of groups for the two independent variables. If values are replicated, the Friedman's test uses the median or mean and replaces all values with it.
Conceptual framework Conceptual model
Figure 1. The conceptual model shows the effects of age and education on employment. Age
and education have an individual effect on employment, but there is also an interaction effect between age and education on employment.
A conceptual framework was made to conceptualize the relationships between the three variables. Age is ranged from 15 to 69 and is expected to have a positive relationship with employment. Therefore if a person gets older he or she gets more employed. Educational attainment level is also expected to have a positive relationship with employment. A persons chance on employment is higher if he/she attained a high education level. Lastly, a positive interaction between age and educational attainment level is expected. This means that the
8 increase in employment when a person is getting older is higher for high educated people than for low educated people.
Hypotheses
The first hypothesis is that the age of a person has a moderate positive effect on the employment of a person. Therefore, in the higher age range group the employment rate will be higher than in the low age range group.
Hypothesis 1a: Age has a positive effect on employment.
The second hypothesis is that the educational attainment level of a person has a positive effect on employment. People with high educational attainment are more likely to be employed than persons with low or medium educational attainment.
Hypothesis 1b: Educational attainment has a positive effect on employment.
It is also expected that the educational attainment level has a positive effect on the positive relation between age and employment. This means that people that attained high education levels will on average stay employed longer than people with low or medium attained education levels.
Hypothesis 2: If the level of educational attainment increases, the positive effect of age on employment increases.
Methodology
In this section the path of database exploration is described, the different variables and their characteristics are mentioned and lastly the statistical analyses are written down. This part clarifies the methods of the research.
Database exploration
To conduct this research, open databases have been searched for useful data. An intensive exploration of multiple online open databanks has been done to find the appropriate
9 data to be able to test the hypotheses and therefore to answer the research question. The appropriateness of the data depends on several factors. Firstly, the dataset must contain data collected in the Netherlands, because this is the research locality. A large dataset is preferred. A larger dataset is more likely to lead to significant results and that in turn influences the impact of the study. It also means that the data are more likely to be a better representation of the population. Another important factor that must be taken into account is if the variables in the dataset are scaled in an appropriate and commonly accepted way. This means that after the data is analyzed the results can be compared to the results of other studies.
To search for potentially useful open data sources keywords were employed. Examples of these keywords are: education, educational attainment level, age, unemployment, employment rate, databank, open data, labour market and Netherlands. With all the above mentioned conditions taken into account multiple databases were found and listed. After a good comparison the choice for the best useful database was made. Below are the databases that were found.
Eurostat. Eurostat1 is the statistical office of the European Union and is situated in Luxembourg. Its task is to provide the European Union with statistics at European level that enable comparisons between countries and regions. The site of Eurostat offers datasets on a lot of different subjects ranging from general and financial statistics to environmental and technological statistics. It also has done data collection for periods longer than ten years, which makes that different time periods can be compared. Anyone who enters the site can freely access the datasets and download them in a variety of format.
1
10 The World Bank Group. The World Bank Group2 is an organization that has the mission to end poverty all around the globe. In order to help people understand the differences between countries worldwide they maintain an open databank. There is free and open access to data about development in countries all over the world sorted by country and by topic. The Netherlands is also included. Topics are for example 'Social development', 'Aid effectiveness', 'Education' and 'Urban development'. Data is continually updated for each country.
Quandl. Quandl3 is a data search engine which contains over 9 million datasets from 500 sources, structured and highly usable. It offers data from central banks, exchangers, brokers, statistical agencies, think-tanks, academies and private companies in countries all over the world. Subjects are ranging from education, demography, health and housing to stocks, economics and markets. All data is free to access and available in a variety of format.
LABORSTA. LABORSTA4 is an International Labour Office database operated by the International Labour Office Department of Statistics. It contains data and metadata from over 200 countries and territories. Therefore, almost every country in the world is included and all topics are labour related. The database has not been updated since 2008.
ILOSTAT. ILOSTAT5 is the follow up database of LABORSTA. It is also a database of labour statistics that provides multiple datasets with annual and infra-annual labour market statistics for over 100 indicators and 230 countries, areas and territories. However, it only contains data from 2008 until now.
2http://datacatalog.worldbank.org 3http://www.quandl.com 4 http://laborsta.ilo.org 5 www.ilo.org/ilostat
11 After making this list and considering all five data sources, Eurostat, LABORSTA and ILOSTAT had useful datasets. A big difference between the dataset from Eurostat and those from LABORSTA and ILOSTAT was that Eurostat offers data on employment as rates of the total working population instead of absolute numbers of employed people, which makes the possible outcome easier to understand. Another factor that favours Eurostat is that the earlier mentioned dataset contains both age and educational attainment level as predictors of employment. LABORSTA and ILOSTAT has a separate dataset for each predictor. A third reason is that the dataset from Eurostat is bigger and since that is preferred, Eurostat was chosen as the most appropriate data source for this study.
Variables
The dataset used is from Eurostat and it contains three variables. Employment rates from the second quarter of 1998 until the first quarter of 2014 in the Netherlands are listed in a table, sorted by age group and educational attainment level. The dataset was imported in SPSS Statistics 20 and errors were removed.
Age. Age is the independent variable (IV) and is expressed in years. The range is from 15 to 67, because that is the working population in the Netherlands. The variable age is divided in 11 groups of five years (15-19, 20-24, 25-29 etc.). Because the age variable is divided in groups of five years the maximum age group used from the dataset is 65-69.
Employment. The dependent variable (DV) is employment. For this variable the employment rate is used and is expressed as a percentage of the working population. This consists of all persons between 15 and 67 years able to work. Again, because the variable age is divided in groups of five years the maximum age group is 65-69.
Educational attainment. The moderating variable (M) is educational attainment. Because education differs internationally due to differences in philosophies, aims and
12 contents or culture, the United Nations Educational Scientific and Cultural Organization (UNESCO) have given no definition to education. Although, UNESCO does describe it. Education comprises all deliberate and systematic activities designed to meet learning needs, where learning means any improvement in behaviour, information, knowledge,
understanding, attitude, values or skills (UNESCO, 2011).
Worldwide the education systems differ from each other in terms of curricular content and structure. This makes it difficult for national policymakers to make a comparison between their own education systems and those of other countries or to benchmark progress towards national and international goals (UNESCO, 2011). This fact made the UNESCO develop the International Standard Classification of Education (ISCED) to create the ability to compare education statistics and indicators across countries on the basis of uniform and internationally agreed definitions. The ISCED 2011 scale is the scale that is used in the dataset and is more appropriate than the ISCED 1997 scale, because it contains changes in the education systems since the UNESCO revised the scale in 1997.
The ISCED 2011 Scale levels are ranged from 0 to 8. The lowest level is when only early childhood education is attained. The next level means that a person has done primary education and level 2 indicates that lower secondary education is attained. This is simply the first three years in high school. Level 3 is upper secondary education which is the last years in high school for one group and the job preparing education for another group. Level 4 post-secondary non-tertiary education and is job preparing education as well, but on a higher level. The fifth level is the short-cycle tertiary education, which contains education that prepares persons for the highest three levels of education. Level 6 is attained when a person achieves a bachelor in tertiary education. On academic knowledge as well as professional knowledge, skills and competencies. The seventh level is a master's degree. Also here it can be a degree
13 on academic level as well as on a professional level that focuses on achieving skills and competencies for on the job. The last level is getting a doctor's degree. This is an advanced research qualification (UNESCO, 2011).
Table 1
Description of the variables.
Variable Values Description Code
Age 15-69 Age in years. Age is grouped with a five year range (15-19, 20-24.... 65-69) Employment 0 - 100% of working population. Proportion of employed people.
Employment rates are presented between 0-100 with 1 decimal.
Educational attainment
ISCED levels 0 - 8
The education level that is attained using the ISCED 2011 scale.
The first group containing level 0-2, the second group containing level 3 and 4 and the third group containing level 5-8.
Analyses
Moderation means that a variable specifies the conditions under which a given predictor is related to an outcome (Baron & Kenny, 1986). The moderator explains when a dependent variable and an independent variable are related. Because the introduction of a moderating variable changes in which way or how big the relationship between two variables is, moderation implies that there is an interaction effect. Three types of a moderating effect are possible. Moderation could be enhancing, which means that when the moderator increases the effect of the predictor on the outcome increases. Or it can be buffering, which implies that an increase of the moderator decreases the effect of the independent variable on the dependent
14 variable. Thirdly, it can be antagonistic. This is the reversion of the effect of the predictor on the outcome when the moderator increases.
The analysis starts with listing descriptives of the variables to show what the mean level and range of the different variables are. Now frequencies show numbers regarding the amount of employment rates per age group and educational attainment level.
To find statistical evidence supporting the first two hypotheses correlations between the concerned variables are analyzed. Because both age and educational attainment level are ordinal variables a Spearman rank-order correlation is used. Another assumption for this test is that the relation between two variables is monotonic. Therefore a boxplot for both age and educational attainment is presented, before testing the correlation. If the correlations are significant the analysis is proceeded.
Furthermore, a Kruskal-Wallis H test is executed twice. It is performed once for age and employment and again for educational attainment level and employment. The Kruskal-Wallis H test is a rank-based nonparametric test that is used to determine if there are statistically significant differences between two or more groups of an independent variable on a continuous or ordinal variable (Laerd Statistics, 2014). Because our independent variable is categorical, our dependent variable is continuous and the data is not normally distributed the Kruskal-Wallis H test is preferred over a one-way ANOVA.
If significant differences between the groups of both independent variables are found in the previous test we continue the analysis by using a nonparametric Friedman test to detect differences between two or more treatments (the educational attainment levels) across multiple groups (the age groups). In other words, it gives answer to the question whether the means of the different age groups across three educational attainment level 'conditions' are equal. If the answer is no and an overall significant difference in group means is found, it
15 implicates that there is an interaction effect between age and education. The Friedman test is preferred over a two-way ANOVA for the same reason as the Kruskal-Wallis H test is preferred over a one-way ANOVA. From this test, however, nothing can be said about the direction of the interaction effect.
Therefore a multiple hierarchical regression analysis is run to find the possible size and direction of the interaction effect. The regression is run to show the difference in variance explained by the interaction term. If the interaction effect is found to be significant the betas can be looked at to define the size and the direction of the interaction effect. In this way hypothesis 2 can be tested.
Results and discussion
The dataset used is from Eurostat and it contains data collected over a time of sixteen years. For every quarter of a year the employment rate per age group sorted by educational attainment level was collected. In total 3894 employment rates were gathered from the Netherlands from the second quarter of 1998 until the first quarter of 2014.
Descriptives
Due to missing values the balance in the data is a bit shifted. In total there were 57 missing employment rate values and those were excluded. Therefore, a small majority of the employment rates was collected on females with 50,1% (N=1922). Employment rates on males are then 49,9% (N=1915) of the data. The median of age is 40-44 years old.
16 Table 2
Age group frequencies and means of employment.
Frequency Percent Valid Percent Mean of emp. (%)
Y15-19 305 7,9 7,9 56,08 Y20-24 354 9,2 9,2 69,17 Y25-29 354 9,2 9,2 76,38 Y30-34 354 9,2 9,2 77,00 Y35-39 354 9,2 9,2 76,26 Y40-44 354 9,2 9,2 77,27 Y45-49 354 9,2 9,2 77,71 Y50-54 354 9,2 9,2 72,73 Y55-59 354 9,2 9,2 59,96 Y60-64 354 9,2 9,2 30,44 Y65-69 346 9,0 9,0 10,31 Total 3837 100,0 100,0
In Table 2 the frequencies and means of employment per age group are stated. There is a trend that employment goes up between the age of 15-19 until the age of 25-29, where it nearly reaches the maximum value of employment. Then the employment values are steady until the age of 45-49, in which it peaks with a maximum employment rate of 77,71%. After that the employment declines until its lowest level in the age group of 65-69. This is consistent with the earlier research done by Biagi and Lucifaro (2008). The boxplot below visualizes this.
17
Figure 2. Boxplot of the employment rates per age group.
The mean of education is 3,96 and the median of education is ISCED level 3-4. This indicates that the majority of the people have completed upper-secondary education and probably post-secondary non-tertiary education. The frequencies are shown below.
Table 3
Educational attainment frequencies
Frequency Percent Ed0-2 1298 33,8 Ed3-4 1296 33,8 Ed5-8 1243 32,4 Total 3837 100,0
18 The percentages of the education groups are close to each other. The small differences are caused by the earlier mentioned removed values in the dataset. The valid N is the same as for the age variable: 3837.
The range of employment was from a minimum of 0,20% to a maximum of 98,60%. The mean of employment is 62,30% with a standard deviation of 30,41% and a valid N of 3837. On average there were 623 of 1000 people in the working population having a job independent of the age or education level. The boxplot shows the means and the ranges of the employment rates for each educational attainment group. For the ISCED level 0-2 group the mean is 52,30% with a SD of 28,07%. Therefore you can state that the lower educated are with 10% less employed than the average educated people. The ISCED level 3-4 group shows a mean of 64,61% and a standard deviation of 29,98%. Thus, the people belonging to this group were slightly more employed than average. At last, the highest education group of level 5-8 shows that on average 70,35% of the working population is employed with a standard deviation of 30,36%. Persons who attained tertiary education and higher are significantly more employed throughout their lives than both lower education categories you can see a trend here of higher education levels having higher employment rates. This is in support of the general finding done by Ashenfelter & Ham (1979) that education reduces the incidence of unemployment.
19
Figure 3. Boxplot of the employment rates per educational attainment level.
Correlations
As you can state from the boxplot in Figure 2 the relationship between age and employment is non-monotonic. A turning point can be seen between the age group 45-49 years and 50-54 years. Therefore the dataset is split into two datasets with one containing all people between 15 and 44 and the second containing all people between 45 and 69. For both datasets the Spearman correlation is calculated and stated below.
Table 4
Spearman correlation between employment and age (range 15-44 years old)
Employment Age (15-44)
Spearman's rho
Employment Correlation Coefficient 1,000 ,383
**
N 2075 2075
Age (15-44) Correlation Coefficient ,383
**
1,000
N 2075 2075
20 The correlation between employment and age between 15 and 44 is 0,383 and is significant at the p<0,01 level. This indicates that there is a moderate positive effect of age on employment until the age of 44.
Table 5
Spearman correlation between employment and age (range 45-69 years old)
Employment Age (45-69)
Spearman's rho
Employment Correlation Coefficient 1,000 -,753
**
N 1762 1762
Age (45-69) Correlation Coefficient -,753
**
1,000
N 1762 1762
**. Correlation is significant at the 0.01 level (2-tailed).
The correlation between employment and age between 45 and 69 is -0,753 and is significant at the p<0,01 level. This indicates that there is a negative effect of age between 45 and 69 on employment.
The Spearman correlations tell us that, as in line with the first hypothesis, age positively influences employment, but only up to a certain level. Between the age of 15 and 44 employment rates increase proportionally (r = 0,383**). After the age of 45, however, employment rates decline and the effect of age on employment becomes strongly negative (r = -0,753**). Although hypothesis 1a is supported by the first correlation, it is contradicted by the second correlation. Therefore the Spearman rank-order correlations show two
contradicting values for the effect of age on employment and hypothesis 1a is not supported. To tell if educational attainment level influences employment in a positive manner and therefore the second hypothesis is supported another Spearman correlation is calculated. In Figure 3 is seen that the relation between educational attainment level and employment is monotonic and therefore the assumption for a Spearman correlation is met.
21 Table 6
Spearman correlation between employment and educational attainment level.
Employment Education
Spearman's rho
Employment Correlation Coefficient 1,000 ,349
**
N 3837 3837
Education Correlation Coefficient ,349
**
1,000
N 3837 3837
**. Correlation is significant at the 0.01 level (2-tailed).
The Spearman correlation between employment and educational attainment level is 0.349 and is significant at the p<0.01 level. The correlation shows a small positive effect of education on employment. This implies that persons in the higher educational attainment level groups are on average more employed than people in the lower educational attainment level groups. Therefore hypothesis 1b is supported and can be stated that educational attainment does have a positive effect on employment.
Kruskal-Wallis H test
Because the distributions for the different groups do not have the same shape the mean ranks are considered in this test instead of the medians. The Kruskal-Wallis H test tells us whether there are statistically significant differences between the groups.
22 Table 7
Mean ranks for employment and age
Age N Mean Rank
Employment 15-19 305 1333,04 20-24 354 1946,61 25-29 354 2511,28 30-34 354 2574,07 35-39 354 2510,78 40-44 354 2534,98 45-49 354 2542,23 50-54 354 2246,76 55-59 354 1614,05 60-64 354 789,01 65-69 346 390,54 Total 3837
In Table 7 the different mean ranks for each age group are shown. As you can see the mean ranks differ widely across the different age groups. In Table 8 the Chi squared, domains of freedom and the significance level are stated. This Kruskal Wallis H test showed there is a significant difference in employment rate between the different age levels, χ2 (10) = 1717,863;
p = 0,000, with mean rank employment rates ranging from 390,54 (age 65-69) to 2574,07
(age 30-34). Now the same test is performed on employment and education. Table 9
Mean ranks for employment and educational attainment level
Education N Mean Rank
Employment ISCED level 0-2 1298 1419,60 ISCED level 3-4 1296 1987,51 ISCED level 5-8 1243 2369,07 Total 3837 Table 8
Kruskal-Wallis H Test Statisticsa,b for age and employment
Employment
Chi-Square 1717,863
Df 10
Asymp. Sig. ,000
a. Kruskal Wallis Test b. Grouping Variable: Age
23 Table 10
Kruskal-Wallis H Test Statisticsa,b for employment and education
Employment
Chi-Square 473,916
Df 2
Asymp. Sig. ,000
a. Kruskal Wallis Test
b. Grouping Variable: Educational attainment level
Also the Kruskal-Wallis H test for employment and education showed that there is a statistically significant difference between the groups, χ2 (2) = 473,916; p = 0,000, with a mean rank employment rate of 1419,60 for ISCED level 0-2; 1987,51 for ISCED level 3-4 and 2369,07 for ISCED level 5-8. Because the employment rate differences between the age levels and the employment rate differences between the educational attainment levels are both significant the Friedman test can be executed.
Friedman test
The Friedman test is used to find potential differences between age groups that are ranked by educational attainment level.
Table 11
Descriptive statistics of the Friedman test
N Mean Std. Deviation Minimu m Maximu m Percentiles 25th 50th (Median) 75th ISCED0-2 11 51,9955 20,10214 6,25 66,20 49,1500 60,7000 65,5000 ISCED3-4 11 67,8455 25,41362 8,45 84,75 66,2500 78,7500 83,1500 ISCED5-8 11 74,7318 25,11344 15,20 92,60 62,7000 87,5500 90,4500
24 Table 12
Mean ranks of related groups
Mean Rank
ISCED0-2 1,00
ISCED3-4 2,09
ISCED5-8 2,91
From Table 13 it can be concluded that there was a statistically significant difference in employment between the different educational attainment levels, χ² (2) = 20.182, p = 0.000. This indicates that there is an interaction effect between age and education. Since the result of the Friedman test is significant a moderation analysis can be run to examine the size and direction of the interaction effect.
Moderator analysis
To test the hypothesis that education moderates the relationship between age and employment, a hierarchical multiple regression analysis was conducted. First a regression analysis is performed on the main effects of age and education on employment. Age and education accounted for a significant amount of variance in employment, R²= 0.550,
F(12,3824)=389.18, p < 0.001.
Table 14
Model summary of main effects _________________________________ Model R R Square Adjusted R
Square R Square Change
F Change Sig. F Change
1 ,741a ,550 ,548 ,550 389,175 ,000
Then a second regression analysis was performed with the interaction terms between age and educational attainment level added to the regression model. The interaction terms accounted for a significant proportion of the variance in employment, ΔR²= 0.007,
F(32,3804)=149.33, p < 0.001.
Table 13
Friedman Test statistics
N 11
Chi-Square 20,182
Df 2
Asymp. Sig. ,000
25 Table 15
Model summary of main effects with interaction terms
Model R R Square Adjusted R Square Change Statistics R Square Change F Change Sig. F Change 2 ,746a ,557 ,553 ,557 149,330 ,000
Because the overall interaction effect between age and education only accounts for a very small proportion of the variance in employment hypothesis 2 is not supported. There is not enough statistical evidence to state that attainment of higher education levels leads to an increase in the effect of age on employment. Even though the overall interaction effect is neglectable, the individual effects of the different age groups and education levels on employment can be of significant value. Therefore the betas of the different age groups and education levels are considered. The different age categories are coded into 10 variables. The educational attainment level groups are coded into 2 variables. The interaction terms are coded in 20 variables. The first category of educational attainment level is the reference group.
Table 16
Regression coefficients of main effects and interaction effects
Model Unstandardize d Coefficients Standardized Coefficients T Sig. B Beta 1 (Constant) 47,424 25,338 ,000 Age2024 13,790 ,131 5,210 ,000 Age2529 16,214 ,154 6,126 ,000 Age3034 17,072 ,162 6,450 ,000 Age3539 20,581 ,196 7,776 ,000 Age4044 20,991 ,200 7,930 ,000 Age4549 17,627 ,168 6,659 ,000 Age5054 12,090 ,115 4,568 ,000 Age5559 1,364 ,013 ,515 ,606 Age6064 -25,874 -,246 -9,775 ,000 Age6569 -40,244 -,379 -15,204 ,000 ED34 15,264 ,237 5,767 ,000 ED58 12,156 ,187 3,945 ,000 Age2024_ED34 -5,839 -,033 -1,560 ,119
26 Age2529_ED34 3,658 ,021 ,977 ,329 Age3034_ED34 1,971 ,011 ,527 ,599 Age3539_ED34 -6,469 -,037 -1,728 ,084 Age4044_ED34 -3,714 -,021 -,992 ,321 Age4549_ED34 1,074 ,006 ,287 ,774 Age5054_ED34 ,699 ,004 ,187 ,852 Age5559_ED34 -3,742 -,021 -1,000 ,317 Age6064_ED34 -7,445 -,042 -1,989 ,047 Age6569_ED34 -13,623 -,077 -3,632 ,000 Age2024_ED58 2,291 ,013 ,564 ,573 Age2529_ED58 7,163 ,041 1,763 ,078 Age3034_ED58 8,109 ,046 1,996 ,046 Age3539_ED58 3,819 ,022 ,940 ,347 Age4044_ED58 2,862 ,016 ,705 ,481 Age4549_ED58 9,489 ,054 2,336 ,020 Age5054_ED58 11,544 ,066 2,842 ,005 Age5559_ED58 9,843 ,056 2,423 ,015 Age6064_ED58 6,681 ,038 1,645 ,100 Age6569_ED58 -4,191 -,023 -1,026 ,305
In table 16 is shown that only six betas are significant (p < 0.05). These betas range between β= -0.077 to β= 0.066. The attainment of tertiary or a higher education level significantly predicted higher employment rates for people between 30 and 34 and between 45 and 59. Therefore, for people from the age of 45 to 59 attaining a higher education level than post-secondary has the greatest impact on employability. Furthermore, attainment of
upper-secondary or post-upper-secondary education predicted lower employment rates for people between 60 and 69. These found effects are probably the result of coincidence, because there is no logical link between the related age groups and education levels. As Figure 4 shows, the lines of the three different educational attainment levels are fairly parallel shaped, which indicates that there is no interaction effect.
27
Figure 4. Graph of the employment rate means of the different education levels.
Conclusion
In the Netherlands people are forced to work longer by the government, while unemployment rates are rising as well as study costs. Students therefore often choose to avoid high debts by quitting education earlier than wished for and start looking for a job, while earlier studies have shown that higher educational attainment is related to higher employment rates. Human capital theory suggests that attaining higher educational attainment levels increases productivity and therefore the value of an employee to the employer. Educational attainment will then lead to higher employability. From the theory it also follows that if more work experience and skills come with age an older employee is considered as greater human capital. This suggests that in general companies prefer more experienced and older employees. Age was therefore expected to relate positively to employment. Finally, more experienced workers are preferred over less experienced workers. Therefore if young people attain higher education levels the increase in employability was expected to be lower than for
28 older, more experienced people. Shortly, educational attainment moderated the relationship between age and employment.
In this paper an answer was searched for to the question whether in the Netherlands age is related to employment and if the educational attainment level affects this relationship. Evidence was found that age positively influences employment, but only until the age of 44 is reached. After the age of 44 the effect of age on employment turned negative. This is in line with the theory. Another finding was that educational attainment level has a positive effect on employment. Higher educational attainment levels lead to higher employment rates. However, no evidence was found that education moderates the relationship between age and employment. This means that the effect of age on employment was not influenced by attaining higher education levels.
From these results followed that students in the Netherlands nowadays should consider attaining higher education when finishing previous education to avoid becoming unemployed later in their career. Especially people who attained lower-secondary education and lower education levels should try to achieve a higher degree. They are at the greatest risk to lose their job when becoming older. The additional high study costs will probably make up for the greater employment probabilities throughout the career. However, not everybody is able to attain higher education than the level already attained due to a lack of intelligence or money, or because of a bad fit with the educational system.
There were some limitations to consider. One limitation was the fact that industries are not taken into account. Of course, in one industry other factors need to be taken into account when it comes to the retirement age of a person than in other industries. Another limitation was the fact that the findings of this study might not be generalizable across Europe, because working conditions are not the same across countries.
29 A third limitation was that the used data was from an open database and not especially collected for this study. The results might have been different if data was collected especially for this study. The reason was the limited amount of time in which the study had to be finished.
Finally, this paper also contributed to the research on how age relates to employment and the effect education has on this relationship. To our knowledge, no research was done yet on this subject in the Netherlands and it sheds a new light on the importance of education and especially on the effect it has on employment. It is necessary for the Dutch government to deeply investigate the role of education in the quest of employment in such times of high unemployment.
Recommendations for future research
In this study the effect of educational attainment level in general on employment is discussed, but the study does not take into account when people attained this education. An interesting research could therefore be performed on the influence of on-the-job education on employment in different age categories.
Another interesting subject for future research is for which job industries the influence of the educational attainment level on employment is the highest. This could help students to make the decision between leaving education, because for the future industry they will be working in the attained education level matters less than other employee characteristics, or attaining the highest education level possible, because it provides them with certainty about a future job.
A last recommendation for future research is to find whether the youth unemployment is influenced by the educational attainment level and what the differences in youth
30 unemployment among the different education groups are. Do smart adolescents lshow lower employment rates, because they study harder for a bright future?
References
Ashenfelter, O., & Ham, J. (1979). Education, unemployment, and earnings. The Journal of
Political Economy, S99-S116.
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of
personality and social psychology, 51(6), 1173.
Biagi, F., & Lucifora, C. (2008). Demographic and education effects on unemployment in Europe. Labour Economics, 15(5), 1076-1101.
Centraal Bureau Statistiek, (2014). Retrieved July 25, 2014, from http://www.cbs.nl/nl-NL/menu/themas/arbeid-sociale-zekerheid/cijfers/werkloosheid/default.htm Deeltijdwerk in Europa neemt toe, Nederland blijft koploper (2014). Retrieved from
http://www.cbs.nl/nl-NL/menu/themas/arbeid-sociale-zekerheid/publicaties/artikelen/archief/2014/2014-eu-meetlat-deeltijd-art.htm
Devine, T. J., & Kiefer, N. M. (1991). Empirical labor economics: the search approach. OUP Catalogue.
European Commission (2014). Retrieved July 18, 2014, from http://ec.europa.eu/social/main.jsp?catId=1036
Kettunen, J. (1997). Education and unemployment duration. Economics of Education Review, 16(2), 163-170.
Kiefer, N. M. (1985). Evidence on the role of education in labor turnover. Journal of Human Resources, 445-452.
31 https://statistics.laerd.com/spss-tutorials/two-way-anova-using-spss-statistics-2.php Laerd Statistics (2014). Kruskal-Wallis H test. Retrieved from
https://statistics.laerd.com/spss-tutorials/kruskal-wallis-h-test-using-spss-statistics.php McKenna, C. J. (1996). Education and the Distribution of Unemployment.European Journal
of Political Economy, 12(1), 113-132.
Nickell, S. (1979). Education and lifetime patterns of unemployment. The Journal of Political Economy, S117-S131.
Olaniyan, D. A., & Okemakinde, T. (2008). Human capital theory: Implications for educational development. Pakistan Journal of Social Sciences, 5(5), 479-483. Rijksoverheid, (2014). Retrieved July 18, 2014, from
http://www.rijksoverheid.nl/onderwerpen/pensioen
UNESCO, (2011). International Standard Classification of Education 2011. Retrieved from http://www.uis.unesco.org/Education/Documents/isced-2011-en.pdf
Verhoef, F. (2014, May). Gevolgen leenstelsel: Is het nog wel verstandig om te studeren? HP
de Tijd. Retrieved from
