Over the last couple of years, each weekend I have registering how many vacancies websites claim to have. This post shows some of the observations one may draw from the plots.

Data

Data is from general and more specialized websites. The first observations started in 2010. Since that time there were a number of changes. In general, the most useful sites are uitzendbureau.nl, nuwerk.nl nationalevacaturebank.nl, indeed.nl and jobbird.nl. These all cover the whole market.
Uitzendbureau (employment agency) will cater mostly for the short term solutions. As an employer, wants an employee quickly, without much hassle, which can let go just as easy, an uitzendbureau is the port of call. The website covers many of these agencies, from what I have seen, I can confirm the bigger agencies. This makes the number of vacancies they have a marker from the general work market. The other three cover more the jobs at employers themselves. I seem to remember that nationalevacaturebank (national vacancies bank) did a lot of advertisement at the time. Nuwerk.nl is a subsite of nu.nl, a popular news site in the Netherlands. The others I added later to the collection.
In addition, there is VK banen, Intermediair and Stepstone. VK banen (Volkskrant) and Intermediair are the traditional sites for higher educated employees. Back in the time when internet was not a factor of importance, Intermediair came for free for high educated people starting just before graduation and ending at age 40 or 45. Thus traditionally it was one of the best places for vacancies. Volkskrant had somewhat lesser amount of vacancies, and many in education. At some point within the data they merged their vacancy activity. Stepstone is an international entity, which tries to get hold in the Dutch market.
Finally, BCFJobs is a website which caters Bio (life sciences), Chemistry and Food jobs. It does jobs from medium level upwards. It got into this list, because this covers my personal background.

Data import

Over the years I have used several (spreadsheet) programs to store the data. Currently it sits in Libreoffice, thus I will extract the data from an ODS file. read.ods resulted in a list (for each page in the spreadsheet) with an entry with text columns for each column in the page, I did not notice any option to directly employ the column names which were in row 1. Hence there is a bit of processing to obtain a data.frame.

Plots

General observations 

In the first figure there are uitzendbureaus, nuwerk and nationalevacaturebank. The line for nuwerk is terminated. At that point they switched to vacancies supplied by monsterboard. The numbers were totally different. and I stopped registering them.
Uitzendbureaus has quite fluctuating numbers. Especially in 2014, but to a lesser extend in other years, there is a spike just before summer holidays. Much of the work which needs to be done when people go on vacation is supplied via uitzendbureaus. A valley at the end of each year is caused by Christmas vacation. Many business close between Christmas and new year, somewhat depending on the actual weekdays on which they fall. As a result there is a dip in both uitzendbureau and nationalevacaturebank.nl. A final observation, which I find it difficult to explain is that the beginning of just many year shows some optimism. The number of vacancies increases. Then, after the summer vacation, things are more pessimistic. In the crisis years the second half of the year gave an actual decrease, post crisis it is a flat line. One possible explanation may be in the end of the school year just before summer. This means summer and autumn there may be a fresh batch out of school. Most of these may be entering their names at the uitzendbureau, one is required to search for a job, and this is one of the ways to go about that. By the end of the year these will all been supplied. But I could think of other reasons too (e.g. head count low at end of year, optimism/pessimism due to seasons).
Regarding the crisis, starting mid 2011 and continuing till end of 2012 there are less and less vacancies. 2013 is the year things stabilized, while in 2014 things started to get better. In terms of vacancies, we are getting to the level of the beginning of the series, before the crisis hit hard.
Nuwerk was actually able to weather the crisis reasonably well. However, subsequently they did lose a bit of market and the associated uptake in vacancies in 2014 was not captured. I imagine this is part of the reason they started using monsterboard.

Extending the web sites used

In this extended plot there a few extra sources. They are added for completeness sake. In addition, since the number of vacancies covers a larger scale, it is plotted on logarithmic scale. Werk is the government website. On this website one is required to register to which vacancies one has applied. Insufficient activity on the job acquisition will result in cuts in benefits. It also has vacancies. Unfortunately it is not a very good website and quite often in the weekend it is down for maintenance. I gave up on retrying later in the weekend, fortunately I was never in the situation of having to use this website.
Jobbird did quite some advertising at some point. However, there are some odd spikes. I am not sure what that is, but this is for me a reason to doubt if their number of vacancies is a good indicator of what is happening on the jobs market.
Indeed has a number for the new vacancies. This is therefor even more fluctuating. Every school vacation has a bigger or smaller dip. It is also one of the last websites which I added. Especially the large fluctuation was a big reason not to plot these data in the first figure.

Higher education

VK (Volkskrant) was traditionally the website where school teachers were recruited. As a consequence, somewhere end of April, beginning of May, a large number of vacancies appeared for the next school year. These are the peaks visible in the three years for which VK has data. A similar but much less pronounced peak is visible in 2013 and 2014. In the section above, I stated that 2013 was the year things stabilized. This is not true in this plot. Intermediair showed a decrease in 2013. Stepstone was able to gain some market, which they subsequently lost in 2014. Then 2014 was the year things stabilized and 2015 saw increases in jobs. What is happening end of 2015 is something new. Intermediair made a big jump in one week (2125 to 2645 from 12 to 19 September), stepstone followed a few weeks later with some jumps. This really feels like a change in the market.
In contrast, BCFJobs did fairly well through the crisis. However, during and after the summer vacation it did have a loss in vacancies. It is almost as if vacancies have been pulled from BCFJobs and placed into Intermediair and stepstone. As explained above, Intermediair is like the best known place for higher educated people to start their job search, likewise it may be the best place for employers to go to when it gets less easy to find employees. As the more and more people get employed again, this may be forcing the change. But that is speculation.

Conclusion

The number of jobs is fluctuating, depending on vacations, season and progress through the crisis. Regarding the crisis, 2013 was the year things stabilized and 2014 saw an increase in vacancies. For higher educated personnel, this change happened about a year later.

Code

library(readODS)
library(ggplot2)
library(dplyr)
r1 <- read.ods('banen.aantal.clean.ods')
mynames <- sapply(1:11,function(i) r1[[1]][1,i])
mycols <- lapply(1:11,function(i) {
      if (i==1) as.Date(r1[[1]][-1,1],format=’%b %e, %Y’)
      else as.numeric(r1[[1]][-1,i])})
r2 <- as.data.frame(mycols)
mynames
names(r2) <- mynames
r3 <- mutate(r2,
    Intermediair_VK=ifelse(!is.na(VK),NA,Intermediair),
    Intermediair=ifelse(is.na(VK),NA,Intermediair))

l1 <- reshape(r3,
    idvar=’date’,
    direction=’long’,
    varying=list(names(r3)[-1]),
    timevar=’source’,
    v.names=’count’,
    times=names(r3)[-1])

l1 %>% 
    filter(.,source %in% c(‘nationalevacaturebank.nl’,’uitzendbureaus’,
            ‘nuwerk’)) %>%
    ggplot(.,aes(y=count,x=Date,col=source)) +
    geom_line()+
    ylim(0,NA)+
    theme(legend.position=”bottom”)

l1 %>% 
    filter(.,source %in% c(‘nationalevacaturebank.nl’,’uitzendbureaus’,
            ‘indeed’,’werk’,’jobbird’,’nuwerk’),
        Date>as.Date(‘2012-01-01’)) %>%
    ggplot(.,aes(y=count,x=Date,col=source)) +
    geom_line()+
    scale_y_log10()+
    theme(legend.position=”bottom”)+
    labs(y=’Count (log scale)’)

l1 %>% 
    filter(.,source %in% c(‘VK’,’Intermediair’,’stepstone’,
       ‘Intermediair_VK’,’BCFJobs’)) %>%
    ggplot(.,aes(y=count,x=Date,col=source)) +
    geom_line()+
    ylim(0,NA)+
    theme(legend.position=”bottom”)