Status-Based Diff erential Memory and Measurement of Social Capital: Recall Errors and Bias Estimates
Social network research has long emphasized the importance of using longitudinal network data to estimate the causal eff ects of social capital. Lin (2001a) posits that by studying changes in network composition over time, research can more adequately test the direct eff ects of social capital. The main challenge in this endeavor is that the survey instruments used to elicit respondents’ social contacts, such as name generator or position generator, may not have the necessary validity and reliability to detect genuine network changes. Due to recall errors or other measurement problems, the social relationships reported by a respondent ﬂ uctuate from time to time even if the “true” network composition remains stable. Researchers studying network changes would thus have diffi culty distinguishing whether a change in network composition is reﬂ ecting genuine turnover of network members or measurement errors during two panel interviews (Brewer 2000). Without overcoming this measurement instability and better understanding the possible sources of error, it will be impossible to verify propositions related to the eff ects of social capital.