It has been frequently observed that office markets are subject to high volatility in rents and vacancy levels, thus exposing real estate investors to considerable risk regarding expected future income streams. This paper analyzes the determinants of office rents and their variability over time and across submarkets in order to gain additional empirical insights into the rent price formation process. To this end, a hedonic regression framework is presented which yields estimates of rent determinants at various phases of the market cycle using time series data from New York City. A number of location-specific and property-specific variables are identified using the hedonic regression model. Next, tests for structural change are conducted which confirm that rent determinants differ significantly according to the position in the market cycle. Cross-sectional tests for structural change also reveal that rent determinants differ significantly in various areas and submarkets so that no support of a unified rental pricing scheme (i.e. the 'law of one price') is found in the empirical case study. Investigating differences in building quality classes, the estimates show that locational and building-specific quality features are better predictors of rental rates in Class A office buildings than in the lower-grade Class B and C buildings. This is indicative of a higher sensitivity of rental rates in the highly competitive segment of Class A buildings to variations in quality features. Support is also found for the existence of price convergence and spillover effects towards the peak of the market cycle.