Almost 2,5 million tweets were analysed to explore the variation of sentiment over space and time in relation to COVID-19 measures, infection rates and socio-geographical factors in the Netherlands. The sentiment analysis of geotagged and timestamped tweets shows that the mood of people was more negative during the first waves than in later stages. The period containing the strict lockdown and the curfew was not significantly more negative than open periods, which may be attributed to vaccines becoming available. Local COVID-19 infection rates do not explain the great spatial variance in sentiment across Dutch municipalities. However, we found clear evidence for higher urbanisation levels leading to higher negativity rates. Urban residents struggled more during the COVID-19 outbreak compared to people living in smaller towns or in the countryside. We discuss whether this could lead to counterurbanisation.