Selecting neighbourhoods for surveillance

3 January 2023

Investigative journalists have revealed that a number of Dutch municipalities carry out surveillance programmes in specific neighbourhoods, as part of the ‘neighbourhood focused approach’ (WGA) of the national LSI programme. These interventions may involve house searches.

Media attention has focused on the role of algorithms and personal data in the targeting of individual addresses for searches. This is the last step of a three-step selection process:

  1. A municipality joins the national LSI programme;
  2. Within this municipality, one or more neighbourhoods are targeted;
  3. Within these neighbourhoods, individual addresses are targeted to be investigated, which may involve house searches.

This article focuses on the second step. It provides an analysis based on documents obtained by Argos and Lighthouse Reports through FOI requests, and on documents sent to Parliament in response to questions from MP Pieter Omtzigt.

See caveats in the Method section below.

Project plans

Argos and Lighthouse Reports have obtained a number of local WGA project plans. Most contain a description of the targeted neighbourhoods. These descriptions often contain data about demographics and socio-economic characteristics, and sometimes about issues like ‘livability’, crime, and (supposed) welfare fraud.1

To explain why specific neighbourhoods are targeted, project plans describe these neighbourhoods as ‘weak’, ‘vulnerable’, having social problems, being in decline, and having a high share of welfare recipients. Municipalities don’t appear to use formal criteria or procedures to select neighbourhoods.

It might seem an obvious choice to target ‘vulnerable’ neighbourhoods. However, since detecting tax fraud is one of the goals of the WGA, governments involved could also have opted to target rich neighbourhoods.2

Selected neighbourhoods

For 20 municipalities, targeted neighbourhoods could be determined from available documents, and linked with data from Statistics Netherlands. The projects have been carried out over the past twenty years. This section discusses how these targeted neighbourhoods compare to other neighbourhoods.

Targeted neighbourhoods tend to be among the poorest neighbourhoods of the municipality they are part of. In terms of average personal income, they have a median rank of 0.14 on a scale that ranges from close to zero (1/n) for the lowest-income neighbourhood, to 1 for the highest-income neighbourhood.

They also tend to be among the neighbourhoods with the highest share of people with a non-western background, with a median rank of 0.83. The focus on this type of neighbourhoods appears to have increased, with the median rank rising to 0.94 for projects carried out in recent years - an increase that is quite striking.

As an illustration, the charts below show data for three local projects. The first chart shows how neighbourhoods differ from the national average in terms of the average personal income of residents.

The labels show the rank of targeted neighbourhoods (e.g., 2/23 or 0.09 for Schadewijk in Oss). Targeted neighbourhoods are among the poorest neighbourhoods in the municipalities they have been selected from.

The next chart shows the share of people with non-western backgrounds, again compared to the national average. Note that ‘non-western background’ is a controversial category; Statistics Netherlands has announced that it will stop using it.3

Targeted neighbourhoods are among the neighbourhoods with the highest share of people with non-western backgrounds in the municipalities they have been selected from.

Impact on neighbourhoods

In 2019, a WGA project in Rotterdam was canceled following protests by neighbourhood residents, supported by trade union FNV. Many residents where offended by the fact that their neighbourhood had been targeted, which they perceived as a form of institutional distrust: ‘why us? why here?’.

Among the documents published by Argos is a draft revised Data Protection Impact Analysis (DPIA) for local projects. The document focuses on the way individual addresses are targeted, but also contains a discussion of the consequences the programme may have for selected neighbourhoods:

In addition, an LSI programme can contribute to stigmatising the neighbourhood and the residents, with all its consequences, for example for taking out insurance (higher premiums), contracting a loan or obtaining credit.

The likelihood that a programme will have adverse effects is estimated to be high; the impact limited or hard to estimate. The document states that the neighbourhood has been selected on the basis of ‘diverse factors’ and that using the ‘instrument’ (with its negative impact on the neighbourhood) is the outcome of an informed decision.

It appears that the discussion of possible negative impacts of the WGA on neighbourhoods was added to the DPIA after the data protection officer of the Zaanstad municipality questioned the legal and ethical aspects of targeting people based on where they live.

Interestingly, the Zwolle municipality has carried out an LSI programme targeting the entire city instead of a specific neighbourhood, because they wanted to avoid stigmatising specific neighbourhoods.4


As usual on this website, this analysis has an exploratory nature and may contain errors.

Two sources were used to identify neighbourhoods selected for WGA projects: local project documents published by Argos, and WGA projects included in a list of LSI programmes sent to Parliament in response to questions from MP Pieter Omtzigt.

Neighbourhood-level data was obtained from Statistics Netherlands for the year in which the selection decision appears to have been made. The list of projects sent to Parliament only contains end dates; pragmatically the decision was assumed to have been made two years prior. In a few cases, an earlier year was chosen because of data availability.

Note that Statistics Netherlands uses two neighbourhood levels: wijken and buurten (wijken consist of multiple buurten). If a project targeted wijken, the analysis was done at the wijk level; if it targeted buurten, it was done at the buurt level. A project that targets both wijken en buurten was excluded from the analysis.

In addition, a number of projects were excluded because the selected neighbourhoods could not unequivocally be linked to neighbourhoods Statistics Netherlands has data for.

Some municipalities have selected multiple neighbourhoods, either simultaneously or in subsequent projects. Of course, previous selections limit the options that are available when a subsequent targeting decision is made.

The findings of the analysis only apply to projects for which the required data is available.


  1. Note that the concept of welfare fraud is not as straightforward as it may seem. What is described as fraud will often in fact be honest mistakes made by people faced with a complex welfare bureaucracy. 

  2. Cf. a report published in 2016 by the Netherlands Scientific Council for Government Policy. This report discusses whether it would be possible to create fraud risk profiles for neighbourhoods, streets or postcodes: «One might target neighbourhoods in Brabant where the likelihood of detecting cannabis production is fifteen times higher than in average neighbourhoods. Similarly, certain neighbourhoods in [high-income areas] Wassenaar, Blaricum and Amsterdam-Zuid might be designated hotbeds of tax avoidance or even white-collar crime, for that matter.» 

  3. Statistics Netherlands assigns the label non-Western migration background to people if they themselves, or at least one of their parents, were born in a country in Africa, Latin-America or Asia (excluding Indonesia and Japan) or in Turkey. In response to controversy over the category, Statistics Netherlands has decided to replace it. 

  4. LSI verslagen 91 t_m 98

3 January 2023 | Categories: data, privacy, trade union, utrecht