The Politics of Drought Relief - Evidence from Southern India Pt 3

8 Tables and Figures Table 1: Summary Statistics for Block-Year Observations Statistic Declarationbt droughtbt marginc(b)e(t) |marginc(b)e(t) | alignedc(b)e(t) N 14, 043 14, 043 14, 043 14, 043 14, 043 Mean 0.330 0.177 0.011 0.105 0.566 St. Dev. 0.470 0.382 0.147 0.104 0.496 Min 0 0 −0.553 0.0001 0 Max 1 1 0.711 0.711 1 Note: This table reports summary statistics for block-year observations of the dependent and independent variables used in the primary specifications of this paper. Declaration is an indicator that a block b received an official drought declaration in year t while drought is an indicator that the block experienced sufficiently below-average June/July or June-September rainfall. The variable margin is the ruling party’s vote margin in the block’s constituency(ies) c(b) for the election e(t) which precedes year t while aligned is an indicator that the margin was positive, or that the ruling party won the election. Table 2: Drought Declarations and Electoral Competition (1) Panel A: droughtbt |marginc(b)e(t) | Observations (2) Declarationbt 0.245∗∗∗ (0.009) 0.028 (0.046) 14,043 0.191∗∗∗ (0.007) −0.190∗∗∗ (0.032) 14,043 0.184∗∗∗ (0.008) −0.909∗∗∗ (0.302) 14,043 |marginc(b)e(t) | Panel B: 0.148∗∗∗ (0.038) −0.009∗∗∗ (0.002) 14,043 0.582 15 ^ c(b)e(t) | |margin droughtbt Observations R2 First Stage F-Stat Specification Block FE State-Year FE (3) OLS No Yes OLS Yes Yes IV Yes Yes Note: Panel A of this table reports estimation of Equation 4, where the indicator for block b receiving a drought declaration in year t is regressed on the absolute value of the ruling party’s vote margin and an indicator for drought occurrence controlling for state-year fixed effects (OLS specification in column 1) and adding block fixed effects (OLS specification in column 2 and IV specification in column 3). Panel B reports the results of the first stage, using the instrument for the ruling party’s vote margin described in Section 4.1. The Kleibergen-Paap Wald rk F statistic is reported. The Anderson-Rubin Wald test has a p-value of 0.0000. Standard errors are clustered by block. ∗ , ∗∗ , ∗∗∗ denote statistical significance at the 10%, 5%, and 1% levels, respectively. Table 3: Drought Declarations and Ruling-Party Incumbency Effects Declarationbt (1) alignedc(b)e(t) Observations Specification drought State-Year FE −0.028∗∗ (0.012) 14,043 OLS Yes Yes (2) −0.111∗∗∗ (0.012) 3,959 RDD Yes Yes Note: This table reports estimation of Equation 5, where the indicator for block b receiving a drought declaration in year t is regressed on the indicator for alignment to the ruling party, allowing for separate linear trends in the ruling party’s vote margin on either side of the cutoff at a margin of 0. Controls include an indicator for drought occurrence and state-year interacted fixed effects. Column 1 reports the OLS estimation with standard errors clustered by block. Columns 2 reports a regression discontinuity design (RDD) specification using an optimal bandwidth of .039 and a triangular kernel, with robust standard errors (optimal bandwidth estimation based on Calonico et al. (2020)). The table reports the effective number of observations for the RDD specification. Standard errors are clustered by block. ∗ , ∗∗ , ∗∗∗ denote statistical significance at the 10%, 5%, and 1% levels, respectively. Table 4: Robustness to Instrument Construction Baseline Instrument District Leave-Out 100 Kilometer Leave-Out 1999 Election Vote Shares (1) (2) (3) (4) Panel A: droughtbt |marginc(b)e(t) | Declarationbt 0.184∗∗∗ (0.008) −0.909∗∗∗ (0.302) droughtbt Observations R2 First Stage F-Stat Block FE State-Year FE 0.184∗∗∗ (0.008) −0.897∗∗∗ (0.301) 0.186∗∗∗ (0.008) −0.730∗∗ (0.332) 0.147∗∗∗ (0.038) −0.009∗∗∗ (0.002) 14,043 0.582 15 0.158∗∗∗ (0.045) −0.009∗∗∗ (0.002) 14,043 0.582 12 |marginc(b)e(t) | Panel B: \ c(b)e(t) | |margin 0.185∗∗∗ (0.008) −0.882∗∗∗ (0.294) 0.148∗∗∗ (0.038) −0.009∗∗∗ (0.002) 14,043 0.582 15 Yes Yes 0.149∗∗∗ (0.038) −0.009∗∗∗ (0.002) 14,043 0.582 15 Yes Yes Yes Yes Yes Yes Note: Panel A of this table reports IV estimations of Equation 4, where the indicator for block b receiving a drought declaration in year t is regressed on the absolute value of the ruling party’s vote margin and an indicator for drought occurrence. Panel B reports the first stage regression for each IV specification. All specifications include block and state-year fixed effects. Column 1 repeats the estimation with the baseline instrument; column 2 increases the leave-out in the shift factor to exclude the votes for the whole district in which block b is located; column 3 instead increases the leave-out in the shift factor to exclude the votes for all constituencies whose centroid falls within 100 kilometers of the centroid of constituency c(b); and column 4 replaces the 2004 baseline vote shares with 1999 vote shares. The Kleibergen-Paap Wald rk F statistic for the endogenous variable is reported. The Anderson-Rubin Wald test has a p-value of 0.0000 in columns 1 through 3 and 0.0033 in column 4. Standard errors are clustered by block. ∗ , ∗∗ , ∗∗∗ denote statistical significance at the 10%, 5%, and 1% levels, respectively. Yes Yes 0.148∗∗∗ (0.038) −0.009∗∗∗ (0.002) 14,043 15 0.184∗∗∗ (0.008) −0.909∗∗∗ (0.302) Yes Yes 0.159∗∗∗ (0.039) −0.009∗∗∗ (0.002) 13,503 16 0.186∗∗∗ (0.008) −0.918∗∗∗ (0.290) (2) (1) 0.003 (0.010) −0.741∗∗∗ (0.268) Yes Yes 0.145∗∗∗ (0.038) −0.002 (0.002) 14,043 14 |marginc(b)e(t) | 0.146∗∗∗ (0.038) −0.002 (0.002) 14,043 14 Yes Yes (4) Normalized Diff. Vegetation Index Declarationbt 0.012 (0.010) −0.734∗∗∗ (0.271) (3) Continuous Rainfall Yes Yes 0.144∗∗∗ (0.038) −0.003 (0.003) 14,043 14 0.005 (0.009) −0.737∗∗∗ (0.267) (5) Soil Moisture Additional Controls for: Yes Yes 0.142∗∗∗ (0.038) −0.002 (0.003) 13,871 14 0.006 (0.010) −0.719∗∗∗ (0.278) (6) Groundwater Level Yes Yes 0.137∗∗∗ (0.043) −0.004 (0.003) 12,587 10 0.006 (0.010) −0.766∗∗ (0.321) (7) Crop Area Sown Note: Panel A of this table reports IV estimations of Equation 4, where the indicator for block b receiving a drought declaration in year t is regressed on the absolute value of the ruling party’s vote margin and an indicator for drought occurrence. Panel B reports the first stage regression for each IV specification. Column 1 reports the specification with baseline controls: block and state-year fixed effects. Column 2 removes all blocks likely to be considered urban, or whose average night lights in 2008 is above the 95th percentile. Each consecutive column then adds additional environmental controls to the baseline specification: a quadratic in the rainfall ratios for June-September, a quadratic in the rainfall ratio for June/July, an indicator for a 4-week dry spell, and an indicator for a prior-year drought (column 3); the normalized difference vegetation index for each of nine 16-day periods overlapping June-September (column 4); mean monthly soil moisture for June through September (column 5); the district-level groundwater level in May and August (column 6); and district-level ratio of crop area sown in year t to the average crop area sown for 2008 to 2018 (column 7). The Kleibergen-Paap Wald rk F statistic for the endogenous variable is reported. The Anderson-Rubin Wald test has a p-value of 0.0000 in columns 1 through 2, .0008 in column 3, .0010 in column 4, .0009 in column 5, 0.0014 in column 6, and .0029 in column 7. Standard errors are clustered by block. ∗ , ∗∗ , ∗∗∗ denote statistical significance at the 10%, 5%, and 1% levels, respectively. Block FE State-Year FE Observations First Stage F-Stat droughtbt \ c(b)e(t) | |margin Panel B: |marginc(b)e(t) | droughtbt Panel A: Urban Blocks Baseline Specification Excluding: Table 5: Robustness to Additional Controls for National Guidelines Table 6: Robustness to Additional Political Variables (1) Panel A: droughtbt |marginc(b)e(t) | (2) Declarationbt 0.184∗∗∗ (0.008) −0.909∗∗∗ (0.302) HHIc(b)e(t) 0.184∗∗∗ (0.008) −0.953∗∗∗ (0.345) 0.020 (0.017) turnoutc(b)e(t) droughtbt |marginc(b)e(t) | 0.148∗∗∗ (0.038) −0.009∗∗∗ (0.002) HHIc(b)e(t) 0.133∗∗∗ (0.038) −0.008∗∗∗ (0.002) 0.045∗∗∗ (0.004) turnoutc(b)e(t) Observations R2 First Stage F-Stat Block FE State-Year FE 0.184∗∗∗ (0.008) −0.878∗∗∗ (0.293) 0.107 (0.099) Panel B: \ c(b)e(t) | |margin (3) 14,043 0.582 15 14,043 0.606 12 Yes Yes Yes Yes 0.152∗∗∗ (0.039) −0.009∗∗∗ (0.002) 0.102 (0.075) 14,040 0.582 15 Yes Yes Note: Panel A of this table reports IV estimations of Equation 4, where the indicator for block b receiving a drought declaration in year t is regressed on the absolute value of the ruling party’s vote margin and an indicator for drought occurrence. Panel B reports the first stage regression for each IV specification. Column 1 reports the specification with baseline controls: block and state-year fixed effects. Column 2 adds the Herfindahl-Hirschman Index (HHI) of (endogenous) vote shares while column 3 adds voter turnout. The Kleibergen-Paap Wald rk F statistic for the endogenous variable is reported. The Anderson-Rubin Wald test has a p-value of 0.0000 in columns 1 and 3 and 0.0001 in column 2. Standard errors are clustered by block. ∗ , ∗∗ , ∗∗∗ denote statistical significance at the 10%, 5%, and 1% levels, respectively. Table 7: Impact of Drought and Declarations on Agricultural Production IHS(∆N DV Ibt ) (1) Rainbt Rain2bt droughtbt−1 Declarationbt−1 droughtbt−1 × Declarationbt−1 (2) (3) 0.452∗∗∗ (0.055) −0.122∗∗∗ (0.023) 0.012 (0.009) 0.063∗∗∗ (0.009) 0.040∗∗∗ (0.011) 0.193∗∗∗ (0.050) −0.029 (0.019) 0.003 (0.009) 0.058∗∗∗ (0.009) 0.052∗∗∗ (0.011) 0.200∗∗∗ (0.050) −0.033∗ (0.019) 0.004 (0.009) 0.058∗∗∗ (0.009) 0.050∗∗∗ (0.011) 14,037 0.752 14,037 0.764 14,034 0.766 Yes Yes No No Yes Yes Yes Yes Yes Yes Yes Yes Rainbt × Declarationbt−1 Rain2bt × Declarationbt−1 Observations R2 Block FE State-Year FE Env. Variables Political Variables (4) 0.092∗∗ (0.047) −0.006 (0.018) 0.003 (0.009) −0.259∗∗∗ (0.074) 0.053∗∗∗ (0.011) 0.454∗∗∗ (0.114) −0.145∗∗∗ (0.044) 14,034 0.767 Yes Yes Yes Yes Note: This table reports estimations of Equation 6 where the dependent variable is the inversehyperbolic sine (IHS) of the NDVI difference, or the difference between maximum and earlyseason vegetation index, for block b in year t. The independent variables are the ratio of current June-September rainfall to its long-term average (Rain) in block b and year t, an indicator for drought occurrence in year t − 1, and an indicator for receiving an official drought declaration in year t − 1. Column 1 reports the specification with block and state-year fixed effects. Column 2 adds controls for mean monthly soil moisture and the June/July rainfall to its long-term-average (and the squared ratio). Column 3 adds controls for the absolute value of the ruling party’s vote margin; the indicator for being aligned to the state ruling party; a comparable indicator for alignment to the federal ruling coalition; voter turnout; the HHI of vote shares; the number of candidates who ran; and indicators for the political party of the MLA(s). Column 4 incorporates interaction terms between the declaration indicator and quadratic in the contemporanous June-September rainfall ratio. Standard errors are clustered by block. ∗ , ∗∗ , ∗∗∗ denote statistical significance at the 10%, 5%, and 1% levels, respectively. Figure 1: Variation in Drought and Drought Declarations Across Election Cycles (a) Karnataka (b) Telangana (c) Andhra Pradesh Note: This figure plots the share of blocks in each year that had a declaration and drought (black); declaration but no drought (blue); drought but no declaration (red); and declaration data missing (grey). The omitted share is blocks with no drought and no declaration. The indicator for drought is defined in Section 2.1. The tables also display the separate election cycles: 2008-2013, 2014-2017, and 2018 for Karnataka and 2009-2013, 2014-2018, and 2019 for Andhra Pradesh/Telangana. The ruling coalitions in Karnataka were Bharatiya Janata Party (BJP)/Independents (IND), India National Congress (INC), and INC/Janata Dal (Secular) (JD(S))/Bahujan Samaj Party (BSP), respectively. In Andhra Pradesh and Telangana INC was the ruling party followed by Telugu Desam Party (TDP) and Telangana Rashtra Samithi (TRS), respectively; in the most recent election, the YSR Congress Party (YSRCP) came to power in Andhra Pradesh while TRS maintained control of Telangana. Figure 2: Variation in Drought and Declarations by Rainfall Shocks (a) Drought-Affected Blocks (b) Drought Declared Blocks (c) Non-Declared Blocks Note: This figure contains a series of scatter plots where the x-axis is the ratio of June-September rainfall to its long-term average (full-season local rainfall shocks) and the y-axis is the ratio of June/July rainfall to its long-term average (early-season local rainfall shocks). Panel A displays the points which meet rainfall criteria outlined in the national guidelines (and therefore are defined as drought-affected in this paper) in green while points not meeting the criteria are grey; the points in green meet either the criteria that the June-September rainfall ratio is below .75 (vertical black line) or that the June/July rainfall ratio is below .5 (horizontal black line). Panel B then contains only block-year observations where a drought declaration was received while Panel C contains only block-year observations where a declaration was not received. The figures highlight the difference between observations eligible for drought declarations according to the (primary) rainfall criteria provided in the guidelines, Panel A, relative to those that do or do not receive declarations, Panels B and C. Figure 3: Binscatter of Drought Declaration Likelihood Against Ruling Party’s Vote Margin Note: Each point in this figure compares the average likelihood of receiving an official drought declaration for all observations in a bin (y-axis) to the average ruling party’s vote margin for all observations in the bin (each bin contains about 100 block-year observations). The curve then represents a regression of the average declaration likelihood on a fourth order polynomial in the average vote margin (with confidence intervals shaded). Figure 4: Impact of Declarations on Agricultural Production by Prior-Year Rainfall Note: This figure plots coefficient estimates and confidence intervals for an estimation of Equation 6 where the dependent variable is the inverse-hyperbolic sine (IHS) of the NDVI difference, or the difference between maximum and early-season vegetation index, for block b in year t and the indicator for drought has been replaced by indicators for the quintile of the prior-year rainfall shock distribution. The rainfall shock is measured as the ratio of June-September rainfall in the prior-year to its long-term-average. The figure displays the coefficient estimates for the declaration indicator interacted with quintile of the prior-year rainfall distribution and the x-axis provides the range of prior-year rainfall shock values for each quintile. The lowest quintile then displays the impact of declarations in areas affected by a negative rainfall shock (less than 78% of normal rainfall) while the highest quintile contains areas with excessive rainfall (more than 28% above normal). The third quintile represents areas with rainfall around the average (about a ratio of 1). Controls include block and state-year interacted fixed effects; mean monthly soil moisture; the June-September rainfall to its long-term-average (and the squared ratio); the June/July rainfall to its long-term-average (and the squared ratio); the absolute value of the ruling party’s vote margin; the indicator for being aligned to the state ruling party; a comparable indicator for alignment to the federal ruling coalition; voter turnout; the HHI of vote shares; the number of candidates who ran; indicators for years since the last election; and indicators for the political party of the MLA(s). Standard errors are clustered at the block level.