A considerable body of research exists on how office rents, vacancy rates and new supply adjust in response to shocks to occupier demand. Research has settled upon an Error Correction Model (ECM) approach for modelling the dynamics. More recent studies have used panel data (such as Hendershott, Jennen and MacGregor, 2013; Adams and Füss, 2012), but there has been little investigation of either the temporal or the cross sectional variation in the adjustment parameters and why these might vary. Furthermore, the econometric complications from using lagged dependent variables in such models have still to be addressed. We use panel data and dynamic panel estimation techniques for 58 US MSA office markets in this paper and we analyse differences in the parameters found for different locations and time periods, including demand and supply coefficients, implied natural vacancy rates and speeds of adjustment to shocks in fundamental variables. Cross-sectional variations are analysed using variables that depict the characteristics of different locations in terms of their economic activity, urban form and real estate markets.