Using algorithms to optimize the allocation of daycare slots
For many parents in Germany, securing a slot in a daycare facility for their child can be a difficult challenge. Some municipalities in Germany are using software to allocate slots more efficiently and equitably. Our food-for-thought paper outlines the potentials and success factors associated with this kind of common good-oriented use of algorithms.
At the end, there were 31 daycare facilities listed in the little green booklet of Emilia’s parents. It was how they tried to keep track of all the different procedures asked of them – some facilities expected them to check back every four weeks, with others every two months. Some requested that parents do so via email, others by phone. Little inspired optimism: Open door events at the facilities were overshadowed by the long lines of anxious parents being told that starting to look for a slot 18 months before their child was ready for daycare was far too late. Emilia’s parents’ experience is far from unique.
Algorithms can solve allocation problems
More than half of all parents with children under the age of six see problems with the allocation of daycare slots. This is one of the findings of a representative survey conducted by the Institut für Demoskopie Allensbach on behalf of the Bertelsmann Stiftung in November 2020. The situation is also dire for many daycare facility managers as registration management requires considerable time investment on their part. The reason for these problems is obvious: Germany does not have enough daycare slots. Anywhere between 300,000 to 380,000 additional daycare slots for children under the age of three need to be created by 2025. While training additional staff and building new facilities often takes several years, an algorithm-based allocation system could help improve some aspects of the problem in the short term.
In recent years, the district of Steinfurt in North Rhine-Westphalia has used the software KitaMatch to allocate daycare slots. Other cities and municipalities also use the open-source software, which was developed by a research group at the ZEW – Leibniz Centre for European Economic Research in Mannheim in 2017. The software helps to match the preferences of parents for certain facilities with generally decided selection criteria. In a first step, a consistent catalogue is created, depicting which selection criteria are relevant for securing a slot. Afterwards, parents’ wishes are queried. The software then uses this information to help daycare facilities better determine which facility should offer which parents a slot. Daycare facility managers can always adjust things “for special reasons” if necessary. On so-called matching days, all available childcare slots are allocated within a few hours, and parents are then informed of the results. Drawing on discussions held with experts in the field, the food-for-thought paper “Per Algorithmus zum Kitaplatz?” (only in German) presents the findings of our analysis and highlights the potentials and success factors of applying such algorithmic systems-
More efficient, saves time and is fairer
There are three potentials associated with applying the KitaMatch software:
1) Applying algorithms enables a more efficient distribution of scarce childcare slots: For starters, parents’ incentives to make “strategic” decisions is lowered because they choose the childcare center they actually prefer. Secondly, we see a more stable distribution when parents no longer have reasons to reject an offered slot because they know that the software has offered them the best slot possible.
2) The use of the software allows those involved to save time because they no longer have to invest hours over hours in searching for the “right” match. As lengthy follow-up rounds are no longer necessary, parents enjoy greater planning security and daycare facilities benefit from the significant reduction in administrative work.
3) In addition, an algorithm-supported allocation system can increase fairness/ justice. For this, a consistent catalogue of criteria is required that creates transparency in the decision-making process for the slot distribution. If the same criteria apply to all participating daycare facilities, the decisions can be reviewed more easily. This helps to keep personal preferences out of the decision-making process and prevents individual children from being disadvantaged or affected by discriminatory decisions.
“However, applying an algorithmic system is not a silver bullet solution. A complex social problem such as the fair allocation of daycare slots cannot be solved through the use of technology alone,” says Julia Gundlach, co-lead of the Ethics of Algorithms project. Among other things, it is important to research how the use of algorithms influence social segregation in daycare centers. What is clear – even without research – is that even the best algorithm cannot create new childcare slots. Therefore, it is important to invest in new childcare slots while at the same time improving the system of allocation itself.
Ensure participatory processes in shaping the use and application of technology
In cases where the use of algorithms can facilitate improvements, it’s important to focus on the concrete implementation. “After all, every context into which a technical system is integrated already has existing structures and social relationships,” says Julia Gundlach, “and they play a key role in determining whether the use of technology actually leads to positive change.” Drawing on the experience from the application of the KitaMatch software in pilot cities and communities as well as through the “Algo.Rules – Nine Rules for the Design of Algorithmic Systems” framework, we can point to relevant success factors which include:
- Competence-building: Users in the daycare sector should have a basic understanding of technology, but those responsible for developing algorithmic systems should also be sensitized to the social context in which the system is to be applied.
- Awareness-raising: In order to generate acceptance for the use of algorithmic systems in participatory-relevant areas, it is important to communicate proactively and fittingly for each target group.
- Participation: In order to ensure that the actual needs of those affected are incorporated into the process from the outset, these individuals should be involved into the process on a regular basis. It’s also important to identify those individuals who have not yet been involved in the allocation of childcare slots and include them.
These success factors require a high degree of willingness on the part of involved individuals, but they also require favorable structural conditions. The latter requires that political leaders promote the development of expertise in public administration and ensure that solutions to social problems are properly supported and sufficiently resourced.
Support common good-oriented innovation
Using the example of daycare slot allocation, the paper demonstrates how algorithmic systems can help solve social problems. The paper is intended to inspire professionals in the daycare sector to test the software but also to show political decision makers how they can promote innovation aimed at serving the common good. Currently, the application of algorithmic systems is primarily driven by economic interests, which is why we lack tangible positive examples of using algorithms that serve the public interest. This creates blind spots with regard to the common good, and efforts to address important challenges we face in society are not sufficiently solution-oriented. We can, however, change this if we create improved framework conditions and funding opportunities for common good-oriented innovation projects.
Months later, Emilia’s parents can still get upset by the fact that there is no solution to such an obvious problem as the poorly coordinated allocation of daycare slots. Throughout their circle of friends, there are countless reports of mothers’ and fathers’ parental leave being overshadowed by the search for daycare. As for Emilia’s parents, they found a good daycare facility for their daughter through an aquaintance just before their parental leave was about to end. And as soon as they found out that they were pregnant with their second child, they reserved another daycare slot there.