UvA PhD student: “Give parents a say in allocation Amsterdam secondary schools”
Distributing eighth graders across Amsterdam’s secondary schools is a mathematical puzzle for the Osvo, the association of Amsterdam school boards, every year. In recent years, fewer and fewer pupils have been placed at their preferred school. Could this be improved? Mayesha Tasnim wrotes her PhD thesis in the Informatics Institute at the UvA to address this question.
Why is placing pupils with HAVO/VWO-recommendations in Amsterdam secondary schools so complicated?
“There are around 65 secondary schools in Amsterdam, a handful of which are very popular. Because not every primary school pupil can be placed at their preferred school, the Osvo, the association of Amsterdam school boards, uses an algorithm to achieve the best possible distribution.”
“The algorithm used by the Osvo comes from New York. It is a world-famous algorithm; its creators even won the Nobel Prize for it. What makes it so special is that it is strategy-proof. This means that strategic choices do not pay off, so you have to be honest about your preferred schools.”
“But what many people don’t realise is that the New York algorithm is a two-sided algorithm. In other words, it’s not just pupils who rank the schools; the schools also rank the pupils based on their performance. In this way, you feed two sets of rankings into the algorithm, which then produces a specific list. This is not the case in Amsterdam, where only pupils rank the schools. That is what makes the situation in Amsterdam so incredibly complicated from a mathematical point of view. In order to be able to allocate pupils anyway, the Osvo is forced to hold a lottery, in which each pupil is given a lottery number. The downside of this is that a lottery isn’t very efficient.”
Why not?
“Because you have no control over a random draw. It may happen that one pupil wants to go to School A and another to School B, yet both end up at the other school. Also, you can end up very low down in the ranking list. In 2017, this meant that some pupils were placed 18th on their ranking list.”
How exactly does that work?
“Suppose you have very popular schools in your preference list, but you get a bad (high) lottery number. In this case, there’s a good chance you won’t be selected for your 1st choice, but also not for your 2nd, 3rd and 4th choice, because pupils with better lottery numbers who put those schools higher up in their list will get them instead.”
“To prevent this, Osvo has introduced a placement guarantee. Pupils with a HAVO/VWO recommendation list twelve preferred schools and are guaranteed a place. Here’s how the Osvo handles it: after the lottery, the schools at the top of the list of pupils who haven’t been placed in their top 12 are contacted with a request to create additional places. This placement guarantee therefore enables parents to make strategic choices.”
How does that work?
“Imagine you’re a parent and you have to list twelve schools. Often, there are only four you really want to attend. You put those at the top. You then list very popular schools further down, so that you qualify for the placement guarantee. There are blogs and YouTube videos circulating about this strategy. But what many parents don’t realise is that strategic choice also carries risks. Namely, that you might not be allocated to your first-choice school at all.”
“What I’m afraid of is the scenario where more and more parents start using that strategy – because they don’t know the risks – meaning the placement guarantee is no longer sustainable. After all, schools cannot place an infinite number of pupils. In my thesis, I show that if 60 per cent of parents make strategic choices, the placement guarantee fails.”
You’ve also raised this issue with the Osvo. What did they say?
“The Osvo says that the placement guarantee based on equal opportunities is still the best system they can offer.”
Do you have an alternative?
“Not immediately. I’ve compared various algorithms, and there’s the Hungarian algorithm that ensures all pupils can get into their top four preferred schools, but even that algorithm isn’t strategy-proof. That’s why the Osvo was not very enthusiastic about it. The point is, as soon as you add even a single improvement to a random lottery, the algorithm is no longer strategy-proof.”
“One question I ask in my thesis is: how bad is it if an algorithm isn’t strategy-proof? To find out, I conducted a questionnaire among parents in Amsterdam with eight graders, in which I explained how they could make strategic choices and also outlined the risks involved. And it turns out that parents are very conservative when it comes to applying strategies when their child’s future is at stake.”
So a smarter allocation system that doesn’t rule out strategic choices isn’t such a bad thing after all?
“That’s exactly what I think. As long as you communicate openly about which strategies can be used and what the risks are. I also think it helps to give parents a say in this decision-making process. So that parents can help decide: do we want less favourable allocations, as long as it’s fair? Or do we want more efficient allocations, with the risk that parents might start using strategies? If parents can be involved in this decision-making, it might prevent a lot of frustration.”
Mayesha Tasnim will defend her PhD thesis on Friday 4 September at 11.00 am: One-Sided Matching Under Strategic Reporting. Insights from Amsterdam School Choice. The defence will take place in the Aula, Oude Lutherse Kerk, and admission is free.
