This paper investigates assumed capitalisation rates in 3400 DCF valuations of office properties in Stockholm, Gothenburg and Malmˆ during the time period 1998-2004. The study investigates determinants of property-level variation in cap rates and how going in and exit cap rates relate to each other. Exit cap rates are stable over time during the studied period but exhibit substantial variation across properties. Part of this cross-sectional variation can be attributed to the location of the property, part of it is due to other characteristics of the property. Exit cap rates are differentiated between properties of the same type on the same market segment, which shows that valuers apply property level fine-tuning when setting exit cap rates. Properties with low market rent and high long-run vacancy assumption typically have high exit cap rates. Properties in peripheral parts of a city typically have higher exit cap rates than properties in central parts. The implicitly assumed going-in cap rate (defined as assumed net operating income year one divided by estimated market value) follows a similar pattern as the exit cap rate but exhibits more temporary, property-specific variation. Going-in cap rates are strongly influenced by temporary deviations of vacancy rates and rents from assumed ìnormalî levels of vacancy and rent. The difference between going-in and exit cap rates is influenced by assumed short-run growth in net operating income in the way stipulated by theory: high assumed short-run growth is associated with going-in cap rates being lower than exit cap rates.