Erratum: The long-term effects of war exposure on civic engagement (Proceedings of the National Academy of Sciences of the United States of America (2021) 118 (e2015539118) DOI: 10.1073/pnas.2015539118)

The author states, “The dataset compiled by Miguel and Roland (2011) was extensively utilized in the empirical sections of this article. Since the publication of my paper, I have become aware that two variables in the publicly available dataset from Miguel and Roland contain errors. The two variables affected by these errors are latitude and distance to the 17th parallel. After discussing the issue with the creators of the measures, Edward Miguel and Gerard Roland, they have been able to identify that the source of the error comes from assuming an incorrect coordinate reference system. When they created the geographic measures, they mistakenly converted from UTM8 coordinate reference system (the original coordinate reference system of our data points for provinces and districts) to latitude and longitude coordinates on the WGS84 reference ellipsoid. This error resulted in the data points being off by about 2 degrees north of their actual location. I have reestimated the main models and rerun the figures of the paper using the corrected measures of latitude and distance to the 17th parallel using the centroid of the provinces. All models report province-level heteroskedastic clustered-standard errors in parenthesis.” The author affirms that “the original findings of the paper concerning the long-term effects of war exposure on civic engagement and participatory values, as well as the exploration of the pathways of persistence, are substantively unaltered or, in some instances, even strengthened by the corrected data. All the robustness checks to the main models also remain substantively unaltered”. (Figure Presented). (Table Presented). The author notes that, in the SI Appendix, page 49, the following additional text should be included as Appendix 1: “Two control variables in the main models are collinear, specifically, Latitude and South, as evidenced by a VIF above 10 in the second stage regression. However, this collinearity does not affect the instrumental variable (distance to the 17th parallel) or the primary predictor of interest, the log of the number of bombs per square meter. variables is not a major concern for interpreting the main estimates, Table X.1 reports the results after excluding either Latitude or South, which reduces the VIF in the second stage to below 5, from either models. The results show that the main effects are unaltered by excluding either of these two control variables.” The SI Appendix has been corrected online.) The author notes that Supporting Table X.1 should be included. The SI Appendix has been corrected online.