Using the PUMS files of the 2000 U.S. Census and 2005–2007 merged ACS data, we study the metropolitan destination choices made by newly arrived immigrants from seven top source countries between 2000 and 2007. Using a multinomial logit model, we find that the dispersion of new immigrant groups varies by origin, although all groups were subject to 1) the attraction of co-ethnic communities, and 2) the positive effects of labor market conditions in the destination (especially the employment growth rate). However, co-ethnic concentration is much more important than labor market conditions in the destination choice decision, particularly for the poorly educated. Conversely, there is a strong negative effect of co-ethnic concentration on highly educated new immigrants, revealing a dispersed geographical pattern of these highly educated immigrants. We also find the importance of employment share growth rate for highly educated immigrants which suggests that they are more sensitive to the upward employment structure at the destination than a specific job growth rate, and they may not be attracted by a metropolitan area with a low-quality employment structure, despite whether there was an increase in high-level jobs.