Using quarterly U.S. census division data for time period 1975-2006, this paper investigates the dynamic relationships among the house prices of nine divisions (regions): Pacific, Mountain, South Atlantic, Middle Atlantic, New England, East South Central, West South Central, West North Central, and East North Central. Johansen’s ML procedure is applied to shed light on the short-run and long-run components on the error correction model. Furthermore, a symmetric error-correction model is estimated followed by the contemporaneous causality structure that is provided by the directed acyclic graphs. The latter is used as an “input” for estimating the impulse response functions along with the forecast error variance decompositions. The results provide evidence of the presence of large number of cointegration relations between the regional house prices in the US. Moreover, in most cases, West North Central and New England appear to strongly and positively lead the house price changes in most other regions. The statement holds for Middle Atlantic which actually generates negative responses. On the other hand, house prices in East North Central and Mountain are highly influenced by changes in house prices of other regions. These results mostly hold for the dynamic period or from time horizon 0 (contemporaneous) to 35 (8.5 years). Furthermore, the real estate market in the US appears to be mainly led by regions that are influential in many other ways, such as financial, economic, etc.