Reading focus
Syllabus: How Do Cultural Norms and Institutions Organize Behavior?
Separate computational reproduction from later-wave evidence, and compare trust outcomes rather than treating the paper as a single verdict.
Why you’re reading it. Reassess the claim of persistence using new observations and explicit measurement limits.
A question to consider. What follow-up could distinguish a changing institutional environment from an unstable survey measure as an explanation for the weakened associations?
Why this reading
Zanella distinguishes reproducing an analysis on its original data from testing its interpretation with later observations. The article asks whether a historical association with reported trust is stable enough to support a claim about persistent cultural transmission.
Before you read
Repeated cross-sectional surveys sample people at different times; they do not follow the same individuals. OLS is a regression method, while 2SLS is an instrumental-variable procedure with additional causal assumptions. Later survey waves do not contain every original trust question or control variable.
Read for
- Separate the successful reproduction of the original estimates from the extension to later Afrobarometer waves.
- Compare the outcomes and survey waves individually, including results that persist, differences in controls and the interpretation of ordered response categories as numbers.
Consider afterwards
- How would changed samples, changed institutions and unstable measurement imply different explanations for a disappearing association?
- What additional evidence would make a stable cultural disposition a better explanation than a response to current conditions?
Editorial context
Several later-wave estimates weaken, especially for relatives and local councils, while evidence for neighbors remains in the available comparison. Only local-council trust is available in the latest waves examined. Some district controls could not be reproduced in later data, and matching ethnic groups does not turn the surveys into a panel of identical people. These limits matter when judging the author’s challenge to the original interpretation. The study does not establish that every historical effect vanishes or that all trust measures are useless. Its materials permit computational reproduction, which this editorial review has not independently rerun.