Assessing livability changes in priority neighbourhoods by Propensity Score Matching
This report investigates whether the trend in livability is more positive in priority neighbourhoods
compared to other neighbourhoods in the Netherlands. We also assess the method by which this is
investigated: Propensity Score Matching. We explain what this method entails and what the added value
of this method is compared to more standard evaluations. However, we also point out what the
limitations are of this method.
The Dutch ‘empowered neighbourhoods’ (krachtwijken) policy
The Netherlands has a tradition of area-based policy, which initially focused on physical improvement
(urban regeneration in the ‘70s and ‘80s) and later also focused on solving social problems (social
renewal policy in the ‘90s and the beginning of the 00’s) (Permentier, Kullberg & Van Noije, 2013). In
2007, the Minister for Housing, Communities and Integration, Elle Vogelaar, announced that her policy
on district improvement would be focusing on only a select number of areas (Bolt & Van Kempen, 2013).
She introduced the Action Plan for Empowered Neighbourhoods. The central government identified 40
priority neighbourhoods in 18 Dutch cities, which dangled at the bottom of the ‘bad lists’ (Huisman,
2013). These neighbourhoods (which cover 83 four-digit postcode areas) had to do with “an
accumulation of problems in relation to unemployment, livability and safety in combination with an
ageing and one-sides housing stock. Specific policy was developed for each of these neighbourhoods to
improve the situation” (Permentier, Kullberg & Van Noije, 2013, p. 111). On the basis of the efficacy of
this policy, other countries can decide on whether to implement it as well or not. Therefore, the
question arises whether there were improvements in the livability of these appointed ‘priority’
neighbourhoods compared to non-priority neighbourhoods in the Netherlands.
The Livability Index
On the basis of the Livability Index, we compared the trend in livability of the priority neighbourhoods to
non-priority neighbourhoods in the Netherlands. The Livability Index was developed by the research
agencies RIGO Research and Advies BV and Stichting Atlast voor Gemeenten. It is an instrument to
measure livability, which integrates two models. The first model uses a range of factors to estimate the
market value of real estate in a neighbourhood. They assume that this market value reflects revealed
preferences and thus livability. The second model uses a range of factors to estimate residents’ views of
the neighbourhood. They assume that a higher satisfaction indicates a higher livability. Each model
accounts for fifty percent of the total score of the Livability Index. Fifty variables, subdivided in five
dimensions, make up the weighted score of the index. The five dimensions are:
1. Housing stock
2. Resident composition
3. Facilities
4. Safety
5. Physical environment
According to Uitermark, Hochstenbach and Van Gent (2017), using the Livability Index has certain
limitations. The index does for example not take into account co-variances, spatial autocorrelations or
multicollinearities among the variables in a model. Besides, there are methodological problems: there is
a lack of transparency and the Livability Index’s data sources and calculating methods have changed
over time, which makes a diachronic comparison of scores problematic. Besides, Uitermark (2003) notes
that livability is a slippery concept. The term can refer to practically every element of the living
environment. The characteristics of a neighbourhood are indicators, but also the socio-economic status
This report investigates whether the trend in livability is more positive in priority neighbourhoods
compared to other neighbourhoods in the Netherlands. We also assess the method by which this is
investigated: Propensity Score Matching. We explain what this method entails and what the added value
of this method is compared to more standard evaluations. However, we also point out what the
limitations are of this method.
The Dutch ‘empowered neighbourhoods’ (krachtwijken) policy
The Netherlands has a tradition of area-based policy, which initially focused on physical improvement
(urban regeneration in the ‘70s and ‘80s) and later also focused on solving social problems (social
renewal policy in the ‘90s and the beginning of the 00’s) (Permentier, Kullberg & Van Noije, 2013). In
2007, the Minister for Housing, Communities and Integration, Elle Vogelaar, announced that her policy
on district improvement would be focusing on only a select number of areas (Bolt & Van Kempen, 2013).
She introduced the Action Plan for Empowered Neighbourhoods. The central government identified 40
priority neighbourhoods in 18 Dutch cities, which dangled at the bottom of the ‘bad lists’ (Huisman,
2013). These neighbourhoods (which cover 83 four-digit postcode areas) had to do with “an
accumulation of problems in relation to unemployment, livability and safety in combination with an
ageing and one-sides housing stock. Specific policy was developed for each of these neighbourhoods to
improve the situation” (Permentier, Kullberg & Van Noije, 2013, p. 111). On the basis of the efficacy of
this policy, other countries can decide on whether to implement it as well or not. Therefore, the
question arises whether there were improvements in the livability of these appointed ‘priority’
neighbourhoods compared to non-priority neighbourhoods in the Netherlands.
The Livability Index
On the basis of the Livability Index, we compared the trend in livability of the priority neighbourhoods to
non-priority neighbourhoods in the Netherlands. The Livability Index was developed by the research
agencies RIGO Research and Advies BV and Stichting Atlast voor Gemeenten. It is an instrument to
measure livability, which integrates two models. The first model uses a range of factors to estimate the
market value of real estate in a neighbourhood. They assume that this market value reflects revealed
preferences and thus livability. The second model uses a range of factors to estimate residents’ views of
the neighbourhood. They assume that a higher satisfaction indicates a higher livability. Each model
accounts for fifty percent of the total score of the Livability Index. Fifty variables, subdivided in five
dimensions, make up the weighted score of the index. The five dimensions are:
1. Housing stock
2. Resident composition
3. Facilities
4. Safety
5. Physical environment
According to Uitermark, Hochstenbach and Van Gent (2017), using the Livability Index has certain
limitations. The index does for example not take into account co-variances, spatial autocorrelations or
multicollinearities among the variables in a model. Besides, there are methodological problems: there is
a lack of transparency and the Livability Index’s data sources and calculating methods have changed
over time, which makes a diachronic comparison of scores problematic. Besides, Uitermark (2003) notes
that livability is a slippery concept. The term can refer to practically every element of the living
environment. The characteristics of a neighbourhood are indicators, but also the socio-economic status
