This paper investigates the empirical validity of the setup of a large-scale government neighbourhood investment programme in the Netherlands. Selection of neighbourhoods into the programme was determined by a measure of neighbourhood quality. At first sight this is a textbook example for the application of a regression discontinuity design to estimate the causal effect of the programme on neighbourhood outcomes. Neighbourhoods close to the threshold should be similar before the programme starts. However, at the discontinuity threshold we observe a surprisingly large gap in the share of non-Western immigrants between neighbourhoods that were selected into the programme and neighbourhoods that were not. In addition, there is non-compliance with the assignment rule based on the quality index. The pattern of non-compliance is consistent with investing in neighbourhoods with a high share of non-Western immigrants. Finally, the way in which neighbourhoods were defined could be a likely explanation for the imbalance in the share of non-Western immigrants at the discontinuity threshold.