SEARCH

SEARCH BY CITATION

1A & 1B: Shows that measurement error in IBC-based civilian casualty data is unlikely to be nonrandom with respect to levels of insurgent violence.

1C: Provides descriptive statistics for the full country and Sunni, mixed, and Shiite areas.

2A: Shows core results are robust to controlling for preexisting trends in attacks and district FE to pick up predictable heterogeneity in trends.

2B: Shows core results robust to dropping Baghdad.

2C: Shows placebo test on core results.

2D: Shows results of trying to predict civilian casualties with leads of SIGACTs.

2E: Shows core results are not present if difference between lagged attacks and average over t to t+3 is placed on LHS.

2F: Shows core results are stronger in areas with more than the median proportion of their population (48.5%) living in urban areas.

2G: Shows core results for different kinds of insurgent attacks.

2H: Shows core results on insurgent killings are robust to population-weighting districts. Coalition results become statistically weaker.

2I: Shows core results on insurgent killings are robust to using the log of casualties on the RHS. Coalition results become statistically weaker.

2J: Shows core results in the full regression (column 5) are robust to including the count of incidents by each party on the RHS.

2K: Shows core results in the full regression (column 5) are robust to allowing a mean shift for district-weeks in which civilians are killed.

2L: Shows core results on insurgent killings are robust to including spatial lag of incidents on the RHS. Coalition results become statistically weaker.

2M: Shows core results are robust to allowing mean shift for any week that includes the first day of the month (to which we attribute killings identified through morgue reports).

2N: Shows core results on Coalition killings are robust to dropping the 7.6% of incidents involving both Coalition and insurgent killings. Insurgent results become statistically weaker.

2O: Shows difference between Sunni and mixed districts in Table 4 is robust to dropping the 7.6% of incidents involving both Coalition and insurgent killings.

2P: Shows core results with Coalition and insurgent killings per 100,000 on RHS.

3: Shows the impact of population density and urbanity on civilian casualty ratios.

4: Shows the impact of CERP projects and spending on civilian casualty ratios.

5: Shows effects of a one-SD increase in civilian casualties on rate of insurgent attacks in different periods.

Figure 5A shows an alternative matching solution to that described in the text.

FilenameFormatSizeDescription
AJPS_542_sm_suppmat.doc584KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.