of tax reforms and elections Figure 2: Probability of tax reforms around elections (percent) Figure 3: Change in the probability of tax reforms (in pp) around elections Figure 4: Change in the probability of tax reforms (in pp) around elections, 12-month windows Figure 5: ROC curve Figure 6: Different tax measures Figure 7: Different tax types TABLES Table 1: Main results Table 2: Robustness Table 3: The role of other factors Table A.1: Elections in the sample Table A.2: The number of observations with reforms in each tax measure and type
probability, using the benchmark model previously estimated. Figure 5 shows how the probability of an ERBS increases as the distance to the next elections shrinks. The figure starts from the highest value of distance to next election in the sample (Nicaragua), holding the other variables in the regression to their sample means. According to Figure 5 , the probability of adopting an ERBS ranges from 76 to 100 percent, depending how close next elections are. Figure 5. Predicted Probabilities with Varying Distance (Months to Next Elections) Sources: Various
distance to the next elections shrinks. The picture starts from the highest value of distance to next elections in the sample (Nicaragua) holding the other variables in the regression to their sample means. According to Figure 5 , the probability of adopting an ERBS ranges from 76 percent to 100 percent depending how close next elections are. Figure 5: Predicted Probabilities with Varying Distance (Months to Next Elections)* * Reserves, openness, political fragmentation, and growth are evaluated at their means Distance (Months to Next Elections) Figure 6
the Electoral College mechanically implements the outcome of a popular vote. Fourth, each election in the sample should be generally regarded as being sufficiently competitive, meaning that there is a real possibility of change in government. Two major criteria apply in this aspect: a) there were no significant concerns before elections that elections will not be free and fair; b) there were no allegations by Western monitors, if any, of significant vote-fraud. For example, although elections in Mexico never resulted in a change of government before 2000, since
coordination problem by influencing the timing of reforms. Appendix Table A.1. Elections in the sample Australia Czechia Germany Japan Poland UK 3/1990 6/1990 12/1990 2/1990 10/1991 4/1992 3/1993 6/1992 10/1994 7/1993 9/1993 5/1997 3/1996 6/1996 9/1998 10/1996 9/1997 6/2001 10/1998 6/1998 9/2002 6/2000 9/2001 5/2005 11/2001 6/2002 9/2005 11/2003 9/2005 5/2010 10/2004 6/2006 9/2009 9/2005 10/2007 5/2015 11/2007 5/2010 9/2013 8
at a predetermined time, and 0 otherwise. The variable ELEENDO i,t equals 1 in country i and year t if an election whose timing was not predetermined took place, and 0 otherwise. The postelection indicators were coded accordingly. Among the 191 elections in the sample, 56.5 percent are classified as predetermined. 14 The baseline regressions are reestimated with the new election indicators. If the baseline results are robust, they should also hold for predetermined elections. The revised model takes the following form: Y i , t = α i + Σ p = 0 2 φ p E L E P