Displacement at the bottom of the Dutch
labor market
An Empirical Analysis
Voorwoord
Deze scriptie is het resultaat en tevens het einde van een lange doctoraalstudie Algemene Economie. Uiteindelijk heeft het hele traject van deze scriptie redelijk lang geduurd, maar toch ben ik blij met het resultaat.
Het idee van de scriptie begon al in september 2006. Toen begon ik met de bachelorscriptie van de studie Economische Geografie. Het resultaat daarvan was een literatuurstudie naar verdringing van laagopgeleiden in Nederland. Die literatuurstudie is in deze scriptie, voor de studie Algemene Economie, uitgebreid met een theoretisch model en empirische toetsing. Het hele traject heeft dus ruim anderhalf jaar geduurd.
Table of Contents
1. Introduction... 4
2. A theoretical model of unemployment among lower educated workers... 7
1. Introduction
Since the 1980s, a main social policy in the European Union is to increase the labor force participation rate and to decrease the unemployment rate. Also the Netherlands defined her main policy as ‘work, work and work’. Starting in 1982 with the so-called Wassenaar agreement, the Netherlands emphasized the relevance of free labor market conditions. This realization initiated successful bargaining rounds between unions, the government and employers’ organizations, resulting in wage rate increases below improvements in productivity. The consequences of these policy measures in the Netherlands are called the ‘Dutch Miracle’, illustrated by a huge decrease in unemployment figures and subsequently a booming economy. Figure 1 presents these unemployment figures and shows that the unemployment rate changed dramatically during the last two decades of the 20th century. Although the Netherlands had an unemployment rate close to
12 percent in 1983, well above the EU average, by the late 1980s the unemployment rate fell to 7 percent and by the end of the 1990s it was as low as 3 percent and well below the EU average (Van Lomwel and Van Ours, 2003; CBS, 2001). Therefore, studying unemployment in the Netherlands in the period just after 1980 is very interesting.
Compared to the total work force, lower educated workers faced significantly higher
unemployment figures than the overall market in the period 1980-2000.1 Figure 1 shows this is
the case. On average, the unemployment rate of the total work force in this period was 6.8 percent, whereas for lower educated workers this was equal to 14 percent, which is more than twice the unemployment rate of the total work force, with a maximum difference of 9.5 percentage points in 1994 (CBS, Statline). Clearly, the figure shows that both unemployment rates do not get nearer to each other. Since the political statement ‘work, work and work’ stands for lower educated workers as for any other educational group, this raises the question why these lower educated workers remained inferior to the rest of the workforce in terms of unemployment rates. Numerous labor economists put forward explanations for the higher unemployment rate of the lower educated between 1980 and 2000, such as lack of job availability for lower educated workers, additional employers’ preferences apart from educational level, for reasons of efficiency, communication skills and flexibility. But mostly mentioned as an explanation for the higher unemployment rate was the rise of the participation rates of new groups in the labor market, in particular of higher educated people, women and students (e.g. Van Ours et al, 1995,
1 Lower educated workers in this paper are defined according to SOI standards. SOI-1 and SOI-2 are
Figure 1: Unemployment rates of the total work force and of lower educated workers
2003; Webbink et al, 1990; Wielers and Van der Meer, 2001). First, lower educated workers may have been displaced by higher educated individuals if these latter were not able to find a job corresponding to their educational level. Since the number of higher educated individuals increased significantly during the last two decades, from 752,000 in 1980 to 1,890,000 in 2000, this might have been a major threat to lower educated workers in the market. Second, since the Wassenaar agreement enforced employers to remove barriers for part-time work, female participation rates in the labor market increased rapidly, because women mostly wanted to have part-time jobs. From 1980 to 2000, the absolute number of women in the labor market increased from 1,372,000 to 2,755,000, indicating a rise in female participation from 28.6 to 52.0 percent (CBS, Statline). At the same time students became more active in the labor market, filling jobs which were generally suitable for lower educated workers. The increased opportunities in the labor market for part-time jobs increased the participation rate of students from 3.7 percent in 1980 to 37.1 percent in 2000, a ten times higher rate (Steijn and Hofman, 2003)2. Steijn and
Hofman (2000) argue that the rise of the participation rates of these three groups, higher educated workers, women and students, weakened the labor market position of lower educated, resulting in displacement and consequently higher unemployment figures for them.
2 Steijn and Hofman (2003) define students as those following secondary, higher vocational and academic
In contrast to these observations, Van Ours and Ridder (1995) state that the observation of diverging unemployment rates among different population groups is not sufficient to conclude that displacement of the lower educated occurred. This is because job competition is not the only factor explaining the difference between unemployment rates of different population groups. Adjustment costs, for example, are generally higher for lower educated workers than for higher educated workers because of their lack of skills. Furthermore, labor market opportunities and availability of jobs are different between the two population groups as well. Technological growth, for example, initiated a shift of employment from the traditional working sectors of the lower educated to the service sectors.
This paper develops a theoretical model of the unemployment rate of lower educated workers and tests it empirically. By doing so, it tries to find out whether the lagging behind of these lower educated workers, in terms of the unemployment rate, was due to the increased participation rates of higher educated workers and women. This is the central question in this paper. In contrast to previous studies in this field (e.g. Steijn and Hofman, 2000; Van Ours et al, 2003; Webbink et al, 1990; Wielers and Van der Meer, 2001), this paper investigates displacement of lower educated workers between 1980 and 2000 empirically.3 In order to
investigate the central question, the unemployment rate of lower educated workers as a dependent variable is regressed on the participation rates of higher educated workers and of women, together with a set of control variables. These control variables are GDP growth over time, a statistic indicating the availability of jobs for the lower educated in traditional sectors, wages per hour, preferences of employers and, finally, the replacement rate.
This paper is set up as follows. First, it presents a theoretical model of unemployment among lower educated workers. In this section, the displacement theories are explained. Also, interpretations of all the control variables in the model are given. Second, the data are presented. Here, it is clarified how the variables are measured and incorporated in the model. Third, the theoretical model is empirically tested. The model, as presented in figure 2 below, is formulated mathematically. Regression analyses are conducted and regression results presented. In section 5, the results are discussed and, finally, conclusions are drawn in section 6.
3 Van Ours and Ridder (1995) also study displacement processes with a self-constructed model. They
2. A theoretical model of unemployment among lower educated workers
Comparing the total workforce with its lower educated counterpart, it is clear that the latter group could not catch up. At the end of the 20th century, lower educated workers had a slightly higher
unemployment rate than in 1980, indicating a relative smaller decline in unemployment after the recession in the early 1990s in comparison to the entire workforce. In short, employment for Figure 2: The theoretical model
Unemployment of Lower Educated
Workers
Displacement GDP Growth Availability of
Jobs Employers’ Preferences
Higher Educated Participation
Female Participation
Student Participation
Sectoral Employment
Wages per Hour
this group was problematic. In this paper a model is built to investigate the reasons for this lagging behind. This section provides a theoretical explanation for the unemployment rate of lower educated workers with the help of figure 2, which shows all the variables used in the model. Although factors, such as GDP growth, availability of jobs, employers’ preferences and the replacement rate, are influencing unemployment as well, the main focus in this paper is on displacement of the lower educated by other population groups.4
2.1 Displacement
The theory about displacement of lower educated workers mentions several types of displacement: displacement by higher educated people, women and students. The basic concept of displacement assumes that high skilled workers who cannot find a job corresponding to their educational level might accept jobs below that level at the expense of lower educated workers. These lower educated workers are bound to search for jobs on an even lower educational level. Finally, at the end of the line, the lowest skilled workers become unemployed due to this displacement process. According to this theory, the number of higher educated workers in the labor market might have affected the unemployment rate of the lower educated, who are at the bottom of the labor market, during the last two decades of the 20th century. To investigate this
possibility, annual figures of higher educated participation rates are added to the model, shown in figure 2. The hypothesis is that a rise in the absolute or relative number of high skilled workers in the labor market increased the chances of displacement of the lower educated and thus increased their probability to become unemployed.
The second type of displacement may be the result of increased female participation. Since the Wassenaar agreement in 1982, unions moderated wage claims and removed obstacles to temporary work. This decline in obstacles to temporary work has stimulated women more to actively participate in the labor market. This rise in female participation increased hiring options for employers, who could now choose between men and women and between full-time and part-time employees. In a displacement perspective, a rise in the female participation rate could have negatively affected the labor market position and opportunities of lower educated workers, since
4 The theory describing unemployment of the lower educated often speaks about a discouraged
most women found jobs which were generally suitable for lower educated workers (Dagevos et al., 1997). This could have been a stimulation for employers to hire more female workers instead of lower educated male workers, as they are more flexible. Lower educated male workers are mostly breadwinners and accordingly not seeking for part-time jobs, since earnings are not sufficient. The hypothesis is that a rise in the net female participation rate reduced labor market chances for lower educated workers.
Theoretically, this may sound convincingly, but many authors found different results. Elhorst (2003) emphasizes the controversy in recent literature on this subject, but concludes from 11 articles that the majority inclines to a negative effect of the female participation rate on the unemployment rate, in contrast to the displacement theory. Although these studies investigated by Elhorst handle female participation rates in combination with the total unemployment rate, whereas this paper deals with the unemployment rate of the lower educated only, it shows that displacement by women is not an undisputed plausible statement. Van Lomwel and Van Ours (2003) conclude that the increase of the number of part-time jobs did not negatively influence the total number of hours worked. Accordingly, it was, at least partly, responsible for a decline in the unemployment rate. The theory behind this is that the growth in part-time jobs also stimulates the growth in full-time jobs, which, in turn, stimulates more people to participate in the labor market. In other words, part-time labor does not merely divide one job over several workers, but creates additional jobs as well. All in all, to conclude that women displaced lower educated workers, the sign of the female participation rate on the dependent variable has to be positive, whereas a negative sign coincides with the majority of recent literature.
‘Dutch miracle’. However, when comparing student labor market figures with those of the total workforce, it can be seen that students formed only a marginal fraction of the labor force. In 1981, 0.25 percent of all workers were students, and this increased to 3 percent at the end of the century (CBS, Statline). Although this is a huge increase, it is questionable whether the increase in student participation in the 1980s and 1990s can be attributed to be a substantial part of the ‘Dutch miracle’. All in all, displacement of lower educated workers by students is possible, although expected to be very small.
Although this paper focuses on displacement of lower educated by increased participation rates of higher educated workers, women and students, it cannot be examined without incorporating some control variables. These variables are also presented in figure 2. The control variables are the GDP growth rate, the availability of jobs as measured by sectoral employment together with wages per hour, employers’ preferences and the replacement rate. All these variables are accounted for, since they are likely to affect the unemployment rate of lower educated. They are discussed in the following sections.
2.2 GDP
2.3 Availability of jobs
The total number of jobs available for lower educated workers is important for their rate of unemployment. If the number of jobs for the lower educated relative to the number of corresponding workers decreases, then the unemployment rate of this group is likely to increase, and vice versa. Therefore, as figure 2 shows, availability of jobs for the lower educated is added to the model. This availability of jobs can be affected in two ways: firstly, by a shift of employment between sectors due to technological growth, and secondly, by the rational decision of employers to hire or lay off employees.
Traditionally, lower educated workers were mostly employed in sectors like agriculture, industry and construction (CBS, 1996; Steijn and Hofman, 2003). These sectors were relatively labor intensive and because the jobs within these sectors required relatively low skills, lower educated workers were attracted. During the 1980s and 1990s, rising investments in all sectors in the economy resulted in technological growth, a shift of production methods and accordingly a change in the demand for employees. More sectors became capital intensive due to these investments, resulting in a lower dependency on working hands. This had several implications. Firstly, the demand for low skilled work decreased due to improved production technologies. Mass production lost its appeal in most sectors and the demand for more differentiated products of high quality increased (Wielers and Van der Meer, 2003). This resulted in decreased opportunities for the lower educated to find a job in their traditional working sectors and induced a shift of employers’ demand for employees from the traditional manufacturing sectors to the service sectors. Secondly, whereas the upgrading of the job structure decreased opportunities of the lower educated, it improved for higher skilled workers. They benefited since more jobs became suitable for them. All in all, technological change induced a shift of demand in favor of employees with higher educated levels, resulting in fewer opportunities for lower educated workers to find or maintain a job in their traditional sectors.
have a positive effect on the dependent variable, indicating a higher unemployment rate with a rise employment shifts.
The second variable affecting the availability of jobs for lower educated workers is wages per hour. Employers decide how many employees they need in their business based most importantly on the cost and benefits of hiring an extra employee. Rising costs compared to their profitability negatively influences the decision of the employer to hire extra workers. Accordingly, costs and benefits of lower educated workers are important determinants for their opportunities to find or retain a job. Thus, it might influence their level of unemployment. Therefore, hourly wages, as a deciding factor of employers’ hiring decision, are incorporated into the model.
2.4 Employers’ preferences
One of the things the Wassenaar agreement in 1982 and following bargaining rounds stimulated was the arrangement by employers and government to reduce barriers for part-time work. The agreement was the foundation for the exchange between lower wages and fewer working hours or more part-time work. This process stimulated women and students more to actively participate in the labor market. Together with the rise in gross participation rates of these population groups, who were willing to work part-time, employers increasingly hired these workers in order to have the option to expand opening hours and fill spots for evening and night work. This resulted in a higher net female participation and an increasingly flexible labor market because employees started to work on different contract bases. Furthermore, besides educational level and flexibility of employees, employers’ preferences towards workers also include their efficiency and a high level of communication skills. This is presented in figure 2. Employers state that students and women are more flexible, have better communication skills and consequently do not need a lot of additional training (Friesen, 1997). Wielers and Van der Meer (2003) agree with this, by saying about the Dutch labor market:
From the perspective of employers, part-time workers are more flexible, reduce labor slack, and therefore increase labor productivity. Other advances for the employers are the relative easiness to dismiss a part-time worker, and the reduced legislation for part-timers. (…). All in all, marginal labor is clearly the least expensive form of employment for the employer.
indicates the relatively weak position of the lower educated in the labor market. Therefore, part-time employment is added to account for employers’ preferences towards flexibility and efficiency. The hypothesis is that a more flexible labor market, where employers choose for part-time employees, reflects worse opportunities for the lower educated.
2.5 Replacement rate
The replacement rate is another control variable added to the model, shown in figure 2, because it affects the searching behaviour of people once they become unemployed. A high replacement rate tends to encourage unemployed people to stay longer in that situation, and gives disincentives to them to apply for a job. A low replacement rate stimulates unemployed people to actively search for a job, since they do not earn much compared to the situation of having a full-time job. Since the 1980s, income differences in the Netherlands have increased by more than 25 percent, indicating a rise in the gap between rich and poor, and in general also between higher and lower educated workers. Social security policy changes in the Netherlands have had a significant role in this, since the government privatized parts of the social security system and disconnected wage growths with social security payments (Goudswaard, De Kam en Sterks, 2000). This resulted in a drop in the replacement rate by almost 15 percentage points in twenty years. Thus, the replacement rate is added to the model because the hypothesis is that it might have changed the gap between higher educated workers and those at the bottom of the labor market in a negative way.
3. Implementation
The dataset used consists of annual figures of all variables from 1980 to 2000, recorded by the Dutch Central Bureau of Statistics (CBS) and the Dutch Central Planning Bureau (CPB). This section explains the variables as they show up in the model.
Figure 3: Unemployment rate of lower educated workers over the total workforce 16 20 24 28 32 36 40 80 82 84 86 88 90 92 94 96 98 00 Unemployment of lower educated over total workforce
compulsory education in this period, leading to a continuous decline in the number of people with hardly any education. The result of the relatively large drop in the number of lower educated workers is that the problem of unemployment for the lower educated in an absolute sense has diminished. Figure 3 shows the decline of the unemployment rate of the lower educated as a percentage of the total workforce. Fewer lower educated workers ended up unemployed, which is positive. On the contrary, in relative terms this is surely not the case. This paper’s model examines whether this is due to displacement or not.
GDP growth is measured as a percentage. Growth is taken because this reflects well the differences in the business cycle between subsequent years, which is needed to distinguish whether the economy is in a recession or not. This variable shows an upward trend, meaning that the economy grew faster at the end than at the beginning of the 1980-2000 period.
The incorporation of the availability of jobs for lower educated workers in the model is done in two ways: firstly, by a statistic introduced by Lilien (1982) describing variances in employment of different sectors in the economy, and secondly, by adding each year’s hourly wages. For the employment differences, this paper identifies two different aggregate sectors, namely the total of traditional working sectors of the lower educated, i.e. agriculture, manufacturing and construction, and non-traditional sectors, such as the service sectors. Lilien tried to explain cyclical unemployment by investigating supply shocks in the economy and fluctuations or shifts from the natural rate. For the latter, he used a variable called sigma, measuring the variance in sectoral employment growth, according to the following formula:
(
)
2 1 1 2 log log −∆ ∆ = = I i t it t it t X X X Xσ
(1)where X denote total employment, subscript i the different sectors in the economy and subscript t the period of time. For our two-sector structure,
σ
can be rewritten as:(
)
(
(
)
)
2 1 2 2 log log log log −∆ + − ∆ − −∆ ∆ = t t t t t t t t t t t X TS X X TS X X TS X TSσ
(2)where TS stands for total employment of lower educated workers in traditional sectors, as explained above. Since the model in this paper only works with two large aggregate sectors, the non-traditional sector in the second part on the right hand side is denoted as total employment minus employment in the traditional sectors. Note that all variables are incorporated in rate form.
Part-time employment, as an indicator of labor market flexibility, is measured as a percentage of the entire working population. A part-time job reflects a job of less than 35 hours a week. The variable has more than doubled from 16.6 percent in 1980 to 41 percent in 2000.
social security system, which resulted in a drop of the replacement rate of nearly 15 percentage points. The expectation is that unemployed people, because of this drop, are increasingly searching for a job. Thus, the expectation of this control variable is that it has a positive sign in the model, meaning that a harsher social security system increases the transition from unemployment to employment.
4. Empirical analysis
The model as it is presented schematically in figure 2 has the following mathematical form:
Ut = Ct + HEPt + FPt + GDPGt + SIGLILIENt + WAGEt + PTEt + REPLRATE t + t (3)
where:
U = unemployment rate of lower educated people (%)
C = constant
HEP = participation rate of higher educated people (%)
FP = female participation rate (%)
GDPG = gross domestic product growth rate (%)
SIGLILIEN = sigma variable meaning shift in industry mix, introduced by Lilien (%)
WAGE = average wage per hour
PTE = part-time employment rate (%)
REPLRATE = replacement rate (%)
, , , , , , = coefficients
= error term
where the error term is a random error with mean zero and variance 2. As the data used in the
regressions are time series data, it is important to test whether the variables are stationary. A stationary variable has a tendency to return to its mean value, has a constant variance, and the covariance between two values from the series depends only on the length of time separating the two values (Hill et al, 2001a).
Figure 4: Regression order
regression results. If the series are non-stationary but have a long-term equilibrium relationship, the variables are cointegrated, implying that they share similar stochastic trends. If cointegration turns out, their relationship can be estimated using least squares regression on levels (Hill et al, 2001a). Furthermore, an error correction model can then be applied. The linkage between cointegration and an error correction model is based on the Granger representation theorem.5 The
advantage of estimating an error correction model is that it opens the opportunity to test for both short-term and long-term effects. The idea is that a proportion of the disequilibrium in one period is corrected in the next period (Engle and Granger, 1987). If series turn out not to be cointegrated, first differences of the series are calculated and the process starts over again by checking stationarity of the first-differenced variables.
Table 1 presents regression estimates of equation (3) using the unemployment rate of lower educated workers as the dependent variable. One way to find out whether the series used in
5 According to this theorem, two or more integrated time series that are cointegrated have an error
Table 1: Regression results
Dependent Variable: U
Method: Least Squares Sample (adjusted): 1980 2000
Included observations: 21 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.946 0.374 2.532 0.025 HEP 0.230 0.676 0.340 0.740 FP -1.443 0.252 -5.735 0.000 GDPG -0.018 0.406 -0.043 0.966 SIGLILIEN 0.200 1.105 0.181 0.859 WAGE 0.021 0.023 0.910 0.379 PTE 0.299 0.363 0.824 0.425 REPLRATE -0.821 0.379 -2.164 0.050
R-squared 0.769 Mean dependent var 0.140
Adjusted R-squared 0.644 S.D. dependent var 0.034
S.E. of regression 0.020 Akaike info criterion -4.704
Sum squared resid 0.005 Schwarz criterion -4.306
Log likelihood 57.393 F-statistic 6.178
Durbin-Watson stat 1.309 Prob(F-statistic) 0.002
this regression are stationary is to test for the absence of a unit root. To test this null hypothesis, an Augmented Dickey-Fuller test is executed in Eviews. Analysis shows that for most series it is not possible to reject the null hypothesis. Apart from the sigma variable for the industry mix, all series contain a unit root, indicating that the regression results, as they are presented in table 1, lack reliability. Consequently, at this stage, it is not correct to conclude anything about displacement of lower educated workers.
The next step is to find out whether the variables exhibit a long-term relationship by testing for cointegration. This is possible since all variables are stationary when measured in first differences. The results of the Engle-Granger test in Eviews on the model’s residuals are shown in table 2. The test statistic tau turns out to be -3.241, which is significantly more negative than the 5 percent critical value of -3.021. Consequently, the null hypothesis of having a unit root in t
Table 2: Cointegration test results Null Hypothesis: t has a unit root
Exogenous: Constant
Lag Length: 0 (Automatic based on SIC, MAXLAG=0)
t-Statistic Prob.*
Augmented Dickey-Fuller test statistic -3.241 0.032
Test critical values: 1% level -3.809
5% level -3.021
10% level -2.650
*MacKinnon (1996) one-sided p-values.
explanatory variables are the female participation rate and the replacement rate. The fact that only two show up statistically significant might be due to the low number of annual observations, since this paper covers the eighties and nineties of the 20th century only.
Before further interpreting the results of the regression as presented in table 1, first, an error correction model is applied. This possibility is the result of the Granger representation theorem, which says that if cointegration turns out, also an error correction model can be applied. For this paper’s model, the error correction model is expressed by:
t t t t t t t t t t t v REPLRATE PTE WAGE SIGLILIEN GDPG FP HEP C U C U + − − − − − − − − + = ∆ − − − − − − − − ) ( 1 7 1 6 1 5 1 4 1 3 1 2 1 1 1
β
β
β
β
β
β
β
α
(4)where the coefficient
α
indicates the speed of adjustment back toward its long-run equilibrium. The hypothesis is that this coefficient is negative, so that a positive (negative) departure from equilibrium in the previous period will be corrected negatively (positively) in the current and future periods. The term between brackets on the right hand side of (4) equals the deviation from the long-run equilibrium in the previous period. This error correction equation reduced to:t t t t C v U = +
ê
+ ∆α
−1 (5)Table 3: Error correction results
Dependent Variable: ∆Ut
Method: Least Squares Sample (adjusted): 1981 2000
Included observations: 20 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.0007 0.004 0.168 0.868
1 −
t
ê
-0.385 0.285 -1.347 0.195The coefficient indicating the speed of adjustment back towards the long-run equilibrium turns out to be -0.385, so that the hypothesis of a negative correcting adjustment parameter can be confirmed. As expected, a positive departure from equilibrium in the previous period will be corrected negatively, and vice versa. This will be done by approximately 38.5 percent each year.
5. Results
This section discusses the regression results as they are presented in table 1. First, this will be done by looking at the traditional displacement theory, which predicts that higher educated workers displace lower educated workers. Table 1 shows that the coefficient of the higher educated participation rate variable is positive, indicating that a relative rise of higher educated workers in the labor market positively affected the unemployment rate of lower educated workers. However, although the coefficient of this variable is positive in the estimate, it is not significant. Accordingly, the hypothesis that higher educated workers displaced lower educated workers in the 1980-2000 period is not statistically confirmed by the data.
variable is significant, it is clear that a positive relation with the dependent variable is absent, meaning that the hypothesis that lower educated workers were displaced by women must be rejected. Whereas some authors argued theoretically that displacement of lower educated workers by higher educated workers or women did occur because of the fewer opportunities of lower educated workers (Wielers and Van der Meer, 2001, 2003; Steijn and Hofman, 2000), this paper shows empirically that they are wrong.
The growth in GDP is negative, as expected, indicating that a recession is responsible for higher unemployment figures, and an economic boom for lower unemployment figures. However, the coefficient is not significant. The sigma variable is positive as predicted in section 2, but insignificant. Accordingly, the hypothesis that the unemployment rate of lower educated workers increased due to a shift in labor demand between sectors cannot be confirmed. The positive coefficient for hourly wages indicates that employers are discouraged to hire employees, which indicates that higher wages tend to increase the unemployment rate of lower educated workers. This variable is insignificant, which indicates that the hypothesis that unemployment of the lower educated is stimulated by rising wages cannot be confirmed. The part-time employment variable is positive, but insignificant as well. Therefore, it is hard to say whether the rise of part-time jobs in the last two decades of the 20th century affected unemployment of the lower educated. The
decline in the replacement rate in the period 1980-2000, as a final control variable, did not negatively affect the unemployment rate of lower educated people. This is in contrast to what was expected. On the contrary, this coefficient shows up statistically significant. The reason for this negative effect on the dependent variable might be that the difference between the unemployment rate of lower educated workers and that of the total workforce increased, as is shown by figure 1. A decreasing replacement rate, as manifested during 1980-2000, hurts lower educated unemployed people more than higher educated unemployed people, because these latter are in general able to find a job on a smaller time scale.
6. Conclusions
last two decades of the 20th century were displaced by women and higher educated workers, the
coefficients of both variables should be positive and significant.
The estimation results of the model showed that the rise of the female participation rate from 30 percent in 1980 to 52 percent in 2000, had a negative rather than positive effect on the unemployment rate of lower educated workers. This implies that the hypothesis that lower educated were displaced by women in the 1980-2000 period in the Netherlands must be rejected. Furthermore, the level of education of the total workforce increased significantly during the testing period. Whereas in 1980, only 15 percent of the total workforce consisted of higher educated workers, this increased to 27 percent in 2000. The theory predicts that this rise in the higher educated participation rate variable worsened the labor market position of lower educated workers because of a possible displacement process. The model in this paper showed that during the 1980s and 1990s, this rise of higher educated people in the labor market affected the dependent variable positively. However, the impact turned out to be insignificant. This means that the hypothesis that a higher number of higher educated people in the labor market displaced lower educated workers cannot be confirmed.
Although this paper captured only the last two decades of the 20th century, it clearly
References
Abraham, K. G. and L. F. Katz (1986) Cyclical unemployment: Sectoral shifts or aggregate disturbances? Journal of Political Economy, vol. 94, pp 507–522.
Blank, R. M. (1994) The dynamics of part-time work. NBER Working Paper Series, working paper no. 4911, 1994/11.
Bulthuis, O. and E. Klok (2004) De effecten van werkende studenten op de Groningse
arbeidsmarkt. RAP Centraal/CAB.
CBS (1996) ‘Tijdreeksen Arbeidsrekeningen 1969-1993: ramingen van het opleidingsniveau, een
tussenstand’. Voorburg/Heerlen.
CBS (2001) Werken en Leren, 2001-2002. Feiten en cijfers over arbeidsmarkt en onderwijs in
Nederland. Voorburg/Heerlen.
CBS (Statline) www.cbs.nl.
CPB Macro Economische Verkenningen, meerdere jaren.
Dagevos, J., L. van der Laan and J. Veenman (1997) Verdringing op de arbeidsmarkt. Van Gorcum & Comp. B.V. Assen.
Elhorst, J. P. (2003) The mystery of regional unemployment differentials: theoretical and empirical explanations. Journal of Economic Surveys, vol. 17, pp 709-748.
Engle, R. F. and C. W. J. Granger (1987) Co-integration and error correction: representation, estimation, and testing. Econometrica, vol. 55, pp 251-276.
Feldman, D. C. (1990) Reconceptualizing the nature and consequences of part-time work. The
Academy of Management Review, vol. 15, pp 103-112.
Friesen, J. (1997) The dynamic demand for part-time and full-time labour. Economica, vol. 64, pp 495-507.
Goudswaard, K. P., C. A. de Kam and C. G. M. Sterks (2000) Sociale zekerheid op het breukvlak
van twee eeuwen. Samson/Kluwer, Alphen a/d Rijn.
Hakim, C. (1998) Social change and innovation in the labor market. Oxford University Press, Oxford.
Hallaire, J. (1968) Part-time employment, its extent and its problems. OECD, Paris.
Hill, R. C., W. E. Griffiths and G. G. Judge (2001a) Undergraduate econometrics. John Wiley & Sons, New York, USA.
Hill, R. C., W. E. Griffiths and G. G. Judge (2001b) Using eviews for undergraduate
econometrics. John Wiley & Sons, New York, USA.
Lilien, D. M. (1982) Sectoral shifts and cyclical unemployment. Journal of Political
Economy, vol. 90, nr 4, pp 777-793.
MacKinnon, J. G. (1996) Numerical distribution functions for unit root and cointegration tests.
Journal of Applied Econometrics, vol. 11(6), pp 601-618.
Martin, J. P. (2000) What works among active labour market policies: evidence from OECD countries’ experiences. OECD Labour Market and Social Policy, Occasional Papers No. 30, OECD Paris.
Martin, J. P. and D. Grubb (2001) What works and for whom: a review of OECD countries´ experiences with active labour market policies. IFAU 2001:14.
OECD (1995; 2001; 2005) Employment outlook. Paris.
Steijn, B. and A. Hofman (2000) ‘Verdringing van laaggeschoolden als beleidsprobleem’, in P. B. Lehning, De beleidsagenda 2000. Strijdpunten op het breukvlak van twee eeuwen. Bussum: Coutinho, pp 206-219.
Steijn, A. J. and W. H. A. Hofman (2003) Students or lower-skilled workers? ‘Displacement’ at the bottom of the labor market. Higher Education, vol. 45, pp 127-146.
Van Lomwel, A. G. C. and J. C. van Ours (2003) On the employment effects of part-time labor.
CentER discussion paper, Tilburg University 2003/04.
Van Ours, J. C. (2003) Has the Dutch Miracle come to an end? CentER discussion paper, Tilburg University 2003/03.
Van Ours, J. C. and S. Nickell (1999) The Netherlands and the United Kingdom: a European unemployment miracle? CentER Discussion Paper, Tilburg University 1999/11.
Van Ours, J. C. and G. Ridder (1995) Job matching and job competition: Are lower educated workers at the back of the job queues? European Economic Review, vol. 39, pp 1717-1731.
Visser, J. (2001) Negotiated flexibility in working time and labour market transitions - the case of the Netherlands. Working paper AIAS, Amsterdam.
Webbink, H. D., C. N. Teulings and P. Tang (1990) De werkgelegenheid voor laagges choolden 1979-1985. WRR (W 45) Den Haag.
Wielers, R. and P. van der Meer (2001) The increased labour market participation of Dutch students. Work, Employment & Society, 15/1, pp 55-71.
