Assessing the Longitudinal Stability of Network Size Using Nsum: Evidence from Chile

The Network Scale-Up Method (NSUM) is widely used to estimate egocentric network size, but its longitudinal stability has been questioned. Variability in survey responses may arise due to recall bias, changing social ties, or methodological inconsistencies. This study investigates the temporal stability of NSUM measures and examines the factors influencing changes in network size, composition, and structure. We specifically compare NSUM outcomes over time across socioeconomic variables (education, class, SES) and gender to understand the robustness of these measures.

Using data from the Chilean Social Longitudinal Survey, which has tracked a national sample of 3000 adults since 2016, we focus on three survey waves (2016, 2018, 2021). The survey includes 25 items estimating the number of acquaintances across occupations, social groups, and named individuals. Person-level, index-level, and item-level analyses were conducted, using descriptive statistics and multilevel models to assess the reliability of the responses.

Our findings reveal moderate inter-wave correlations for egocentric network items (ranging from 0.3 to 0.6), indicating some stability in network reporting. The correlation for the aggregate network size index, particularly McCarty et al.'s Scale-Up estimator, was 0.46, suggesting caution when interpreting longitudinal network data. Regression models indicate that response stability in attitudinal measures strongly predicts changes in reported network size. This points to individual differences in response behavior, such as survey satisficing, as key contributors to the variability in network size estimates.

These insights challenge the assumption that network reporting is inherently stable and highlight the need for careful consideration of survey reliability in longitudinal studies of egocentric networks. The study also provides critical evidence on the effects of survey mode, with 2021 data collected via telephone due to the COVID-19 pandemic.