The CGSS uses face-to-face interview, and the average interview time is about one and a half hours.  Now, the CGSS is testing Computer Aided Personal Interviewing and will use CAPI to replace face-to-face very soon.  The CGSS has a set of strict quality control procedures, which cover pre-fieldwork, in-fieldwork, and after-fieldwork states. In pre-fieldwork stage, all supervisors must receive 40 hours training and finish 4 experimental interviews; all interviewers must receive 25 hours training and finish 3 experimental interviews. In in-fieldwork stage, all interviewers will be accompanied to interview by supervisor at least once. And the finished questionnaires must be 100% on site check and supervisors must do 40% the second day back interviewing. And in post-fieldwork stage, all interviews must 100% mail back interview and 40% telephone interview. In data input and coding stage, the data must be double input and double coding and there are several supervisors to check the double input and coding validation process.

In the Cycle I, the CGSS explored the model to organize large-scale survey project in China. Academic institutions jointly committed the 2003 survey; the 2004 survey was subcontracted to a government agency; and 2005, 2006, 2008 survey were subcontracted to commercial survey companies by public biding.  Comparing these three kinds of implementation models, we could find the survey quality using governmental is the worst.  In fact, it is very difficult to do quality control when using government agency. As for survey company, its fieldwork costs much higher than academic institution’s, and on the contrary, its response rate is lower than using academic institutions. Based on learned in the Cycle I of the CGSS,  the national  Survey Research Center at Renmin University of China (NSRC) organized Chinese Social Survey Network (CSSN) in 2009, 20 universities and 2 provincial social science academies joined CSSN. In 2010 ,the members of CSSN undertook the first annual survey of the CGSS Cycle II in their own provices. As show in figure 4, comparing the response rate, missing value rate and logic error rate for a same set of background variables from 2003 to 2010, we could find the model using academic institutions for fieldwork is the most appropriate for CGSS kind of academic survey projects. After, CSSN  has been undertaking fieldwork for the CGSS, and its members increase to 49 universities by the end of 2014.