Endogenous Taxation in Ongoing Internal Conflict FIGURE 5.Estimated Relationship between Property Tax Revenues and Attacks per Armed Group and Time Period across Colombian Municipalities Guerrilla.1988-1996 Paramilitary,1988-1996 Guerrilla.1997-2002 豆 Paramilitary,1997-2002 Guerrilla,2003-2006 'wno Paramilitary,2003-2006 M//soHu Guerrilla,2007-2010 Paramilitary,2007-2010 'asn -.05 .05 Notes:Regression estimates from Table 2,columns(2),(4),(6),and(8),respectively.Treatment variables in logs as specified in Table 2. (see Civil War Dynamics and Capture in Colombia sec- violence in the period 2003-2006 is associated with a tion).Thus,we would not expect the effects to be as 20%drop and an 8%increase in per capita property strong in the third and especially the fourth period.This tax income over the period 2007-2010 for the case of effectively turns the last period into a placebo.31 guerrilla and paramilitary attacks,respectively(column Because of the log-log specification,estimated coef- 6).32 ficients should be interpreted as the elasticity of per capita property tax revenue with respect to cumula- tive past violence.Given the estimated elasticities,we Robustness can compute substantive effects by multiplying them While our four conflict periods are based on our sub- with changes of interest in the independent variable stantive theoretical and historical understanding of the (expressed in terms of percent changes).Using the evolution of the Colombian civil war,a potential con- most demanding specification in column 2 of Table 2 cern is that the results reported in Table 2 are just an an increase in cumulative per capita guerrilla attacks artifact of an arbitrary aggregation of annual data into (paramilitary attacks)over the period 1988-1996 from these periods.Appendix E shows that this is not the the median to the ninetieth percentile of the distri- case.We rerun our most demanding empirical specifi- bution is associated with an average 37%drop (13% cation (including preperiod tax revenues)using year- increase)in per capita property tax income over the by-year moving windows in which six (eight)years of period 1997-2002.An equivalent increase in cumula- cumulative conflict are correlated with three (five)- tive violence in the period 1997-2002 is associated with year aggregations of tax revenues. a 16%drop and a 11%increase,in per capita prop- Figures E1 and E2 report the results graphically,re- erty tax income over the period 2003-2006 for the case spectively,for the six-and the eight-year moving win- of guerrilla and paramilitary attacks,respectively(col- dows.In both cases,we corroborate the asymmetric umn 4).Further,an equivalent increase in cumulative correlation of guerrilla and paramilitary cumulative past violence of municipal tax performance.Also con- 31Figure 5 is the graphical counterpart of Table2(evencolum). sistent with our main results (Table 2)and with the his- We plot the point estimates as well as the 95%confidence intervals, documenting that the estimated effects decrease over time.The dif- ferences across conflict periods are not statistically significant,with 32 Figure6 reports the marginal-effect plots of the effect of cumu- the exception of the comparison between the first and the fourth pe- lative past guerrilla and paramilitary violence on property tax rev- riod.Again,as expected,during the state resurgence period,point enues,with a horizontal axis support ranging from the median to the estimates lean toward zero. ninetieth percentile. 1007Endogenous Taxation in Ongoing Internal Conflict FIGURE 5. Estimated Relationship between Property Tax Revenues and Attacks per Armed Group and Time Period across Colombian Municipalities Guerrilla, 1988-1996 Paramilitary, 1988-1996 Guerrilla, 1997-2002 Paramilitary, 1997-2002 Guerrilla, 2003-2006 Paramilitary, 2003-2006 Guerrilla, 2007-2010 Paramilitary, 2007-2010 (log) cum. attacks per capita -.05 0 .05 Notes: Regression estimates from Table 2, columns (2), (4), (6), and (8), respectively. Treatment variables in logs as specified in Table 2. (see Civil War Dynamics and Capture in Colombia sec￾tion). Thus, we would not expect the effects to be as strong in the third and especially the fourth period.This effectively turns the last period into a placebo. 31 Because of the log-log specification, estimated coef￾ficients should be interpreted as the elasticity of per capita property tax revenue with respect to cumula￾tive past violence. Given the estimated elasticities, we can compute substantive effects by multiplying them with changes of interest in the independent variable (expressed in terms of percent changes). Using the most demanding specification in column 2 of Table 2, an increase in cumulative per capita guerrilla attacks (paramilitary attacks) over the period 1988–1996 from the median to the ninetieth percentile of the distri￾bution is associated with an average 37% drop (13% increase) in per capita property tax income over the period 1997–2002. An equivalent increase in cumula￾tive violence in the period 1997–2002 is associated with a 16% drop and a 11% increase, in per capita prop￾erty tax income over the period 2003–2006 for the case of guerrilla and paramilitary attacks, respectively (col￾umn 4). Further, an equivalent increase in cumulative 31 Figure 5 is the graphical counterpart of Table 2 (even columns). We plot the point estimates as well as the 95% confidence intervals, documenting that the estimated effects decrease over time. The dif￾ferences across conflict periods are not statistically significant, with the exception of the comparison between the first and the fourth pe￾riod. Again, as expected, during the state resurgence period, point estimates lean toward zero. violence in the period 2003–2006 is associated with a 20% drop and an 8% increase in per capita property tax income over the period 2007–2010 for the case of guerrilla and paramilitary attacks, respectively (column 6).32 Robustness While our four conflict periods are based on our sub￾stantive theoretical and historical understanding of the evolution of the Colombian civil war, a potential con￾cern is that the results reported in Table 2 are just an artifact of an arbitrary aggregation of annual data into these periods. Appendix E shows that this is not the case. We rerun our most demanding empirical specifi￾cation (including preperiod tax revenues) using year￾by-year moving windows in which six (eight) years of cumulative conflict are correlated with three (five)- year aggregations of tax revenues. Figures E1 and E2 report the results graphically, re￾spectively, for the six- and the eight-year moving win￾dows. In both cases, we corroborate the asymmetric correlation of guerrilla and paramilitary cumulative past violence of municipal tax performance. Also con￾sistent with our main results (Table 2) and with the his- 32 Figure 6 reports the marginal-effect plots of the effect of cumu￾lative past guerrilla and paramilitary violence on property tax rev￾enues, with a horizontal axis support ranging from the median to the ninetieth percentile. 1007 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000333