Contents of Web Appendix for
“Electoral Manipulation via Voter-Friendly Spending: Theory and Evidence”
Allan Drazen and Marcela Eslava
July, 2008
This Appendix presents supplementary material for the paper “Electoral Manipulation via Voter-Friendly Spending: Theory and Evidence”
1. Party Correspondences: matches between political groups that participated in elections and the two main parties (Liberal and Conservatives). File: party linkages.xls
2. Sources of Funding: lists the sources of funding for different categories of spending by municipal governments. File: sources of funding.doc
3. Full sets of results for the regressions corresponding to Tables 2 and 3: programs and output files
3.1 arrbond_ac_one_l2insz_PIBti_gab_sin_nbi_07_08.do & resABnceros_one_l2insz_PIBti_gab_sin_nbi_07_08.txt: runs arellano-bond regressions for tables 2 and 2a, column 1. The results shown in table 2 correspond to the 3rd set of controls in the do-file.
3.2 arrbond_ac_one_l2insz_PIBti_sin_nbi_07_08.do & resABnceros_one_l2insz_PIBti_sin_nbi_07_08.txt : runs arellano-bond regressions for table 2 and 2a, column 2. The results shown in table 2 correspond to the 3rd set of controls in the do-file.
3.3 gmminotnol_cont_2002_PIBtiga_sin_nbi.sas & gmminotnol_cont_2002_PIBtiga_sin_nbi.lst: runs gmm with the Anderson-Hsiao matrix of instruments for column 1 of panels I and II of Table 3.
3.4 gmminotnol_cont_2002_PIBti_sin_nbi.sas & gmminotnol_cont_2002_PIBti_sin_nbi.lst: runs gmm with the Anderson-Hsiao matrix of instruments for column 2 of panels I and II of Table 3.
4. Alternative sets of results:
4.1 Effect of elections on the composition of total expenditure: regressions including Unsatisfied Basic Needs as additional control in the regressions. Outputs:
4.1.1 resABnceros_one_l2insz_PIBti_gab.txt: runs arellano-bond regressions, controlling for total spending, uses two different interest rates
4.1.2 resABnceros_one_l2insz_PIBti.txt: runs arellano-bond regressions, not controlling for total spending, uses two different interest rates.
4.2 Effect of elections on different types of expenditures: regressions including Unsatisfied Basic Needs as additional control in the regressions.
4.2.1 gmminotnol_cont_2002_PIBtiga.lst: runs gmm with the Anderson-Hsiao matrix of instruments controlling for total spending, includes unsatisfied basic needs as one of the controls.
4.2.2 gmminotnol_cont_2002_PIBti.lst: runs gmm with the Anderson-Hsiao matrix of instruments not controlling for total spending, includes unsatisfied basic needs as one of the controls.
4.3 Voting model: voting_07_08_todofin_wa.txt runs different versions of the voting model:
· Regressions in Tables 4-6 with the swing dummy reported in the paper (“swing1a”) and with a different definition of the swing dummy, based on the share of votes obtained by each party (“swing2c”). In this version a municipality is non-swing if one of the two parties won at least 50% of votes in every election in the sample.
· Regressions in Tables 5 and 6 that include the swing dummy (swing1a or swing2c) or restrict to a value of this dummy, but restricting sample to municipalities with at least three electoral observations.