I happened to read this right after the article on Simpson's Paradox, and you put your finger on exactly what I was thinking - it's almost certain that there are many hidden variables that predict whether or not you work in an open office, and many of these likely correlate very strongly with job satisfaction, potentially inverting the relationship just based on relative employment rates in different areas. Just knowing that satisfaction correlates with open floorplans doesn't tell us anything solid about whether you should choose one or not.