if _sp == 'no': cost = ''
print(_sp)
plt.plot(sp_files[_sp].sum(level='rp').index,
sp_files[_sp]['event_cost'].sum(level='rp') * to_usd / 1.E6,
label=_sp + cost,
color=pairs_pal[_ix])
plt.plot(sp_files[_sp].sum(level='rp').index,
pds_effects['dw_DELTA_' + _sp].sum(level='rp') * to_usd,
label=_sp + ' benefit',
ls=':',
color=pairs_pal[_ix + 1])
_ix += 2
title_legend_labels(plt, 'Sri Lanka', 'Return Period [years]',
'SP cost when event occurs [M USD]', [1, 1000])
plt.xscale('log')
plt.gca().get_figure().savefig(out + 'sp_costs.pdf', format='pdf')
########################################################################
# Samurdhi is a constant payout, but there is still RP-dependence
# - cost of disbursement (annual & when event occurs)
# - wellbeing benefit (annual & when event occurs)
# - cost-benefit ratio
# x-axis is targeting error
# y-axis is in mUSD
pds_effects = pds_effects.reset_index('rp')
_wid = 0.45
rp_range = [5, 10, 25, 50, 100]
def run_poverty_duration_plot(myC):
# Load file with geographical (region/province/district) as index
df = pd.read_csv('../output_country/' + myC + '/poverty_duration_no.csv')
df = df.reset_index().set_index(df.columns[1])
geo = df.index.name
all_geo = np.array(df[~df.index.duplicated(keep='first')].index)
# used in groupby
df['country'] = myC
# assign deciles
listofdeciles = np.arange(0.10, 1.01, 0.10)
df = df.reset_index().groupby(
'country', sort=True).apply(lambda x: match_percentiles(
x,
perc_with_spline(reshape_data(x.c), reshape_data(x.pcwgt),
listofdeciles), 'decile'))
# Load additional SP runs
_sp = []
for f in glob.glob(
'/Users/brian/Desktop/BANK/hh_resilience_model/output_country/' +
myC + '/poverty_duration_*.csv'):
_ = f.replace(
'/Users/brian/Desktop/BANK/hh_resilience_model/output_country/' +
myC + '/poverty_duration_', '').replace('.csv', '')
_sp.append(_)
for iSP in _sp:
_ = pd.read_csv('../output_country/' + myC + '/poverty_duration_' +
iSP + '.csv')
df[['t_pov_inc' + iSP, 't_pov_cons' + iSP,
't_pov_bool' + iSP]] = _[['t_pov_inc', 't_pov_cons', 't_pov_bool']]
############################
# Do some plotting
#plot_crit = '(t_pov_bool)&(hazard=="PF")&(rp==500)'
#df.loc[df.eval(plot_crit)].plot.hexbin('dk0','t_pov_cons')
#plt.gca().get_figure().savefig('../output_plots/'+myC+'/poverty_duration_hexbin_no.pdf',format='pdf')
#plt.cla()
#df.loc[df.eval(plot_crit)].plot.scatter('dk0','t_pov_cons')
#plt.gca().get_figure().savefig('../output_plots/'+myC+'/poverty_duration_scatter_no.pdf',format='pdf')
#plt.cla()
############################
df = df.reset_index().set_index(['hazard', 'rp', 'decile'])
df_dec = pd.DataFrame(index=df.sum(level=['hazard', 'rp', 'decile']).index)
# Populate the df_dec dataframe now, while its index is set to ['hazard','rp','decile']
# Number of individuals who face income or consumption poverty
df_dec['n_new_pov_inc'] = df.loc[df.t_pov_bool == True, 'pcwgt'].sum(
level=['hazard', 'rp', 'decile'])
df_dec['n_new_pov_cons'] = df.loc[df.t_pov_bool == True, 'pcwgt'].sum(
level=['hazard', 'rp', 'decile'])
# Individuals who face income or consumption poverty as fraction of all individuals
df_dec['frac_new_pov_inc'] = df_dec['n_new_pov_inc'] / df['pcwgt'].sum(
level=['hazard', 'rp', 'decile'])
df_dec['frac_new_pov_cons'] = df_dec['n_new_pov_cons'] / df['pcwgt'].sum(
level=['hazard', 'rp', 'decile'])
# Among people pushed into pov: average time in poverty (months)
for iSP in _sp:
df_dec['t_pov_inc_avg' + iSP] = 12. * (
df.loc[df.eval('t_pov_bool' + iSP + '==True'),
['pcwgt', 't_pov_inc' +
iSP]].prod(axis=1).sum(level=['hazard', 'rp', 'decile']) /
df.loc[df.eval('t_pov_bool' + iSP + '==True'),
'pcwgt'].sum(level=['hazard', 'rp', 'decile']))
df_dec['t_pov_cons_avg' + iSP] = 12. * (
df.loc[df.eval('t_pov_bool' + iSP + '==True'),
['pcwgt', 't_pov_cons' +
iSP]].prod(axis=1).sum(level=['hazard', 'rp', 'decile']) /
df.loc[df.eval('t_pov_bool' + iSP + '==True'),
'pcwgt'].sum(level=['hazard', 'rp', 'decile']))
for iloc in all_geo:
df_dec['t_pov_inc_avg_' + iloc] = 12. * (
df.loc[df.eval('(t_pov_bool==True)&(' + geo + '==@iloc)'),
['pcwgt', 't_pov_inc']].prod(axis=1).sum(
level=['hazard', 'rp', 'decile']) /
df.loc[df.eval('(t_pov_bool==True)&(' + geo + '==@iloc)'),
'pcwgt'].sum(level=['hazard', 'rp', 'decile']))
df_dec['t_pov_cons_avg_' + iloc] = 12. * (
df.loc[df.eval('(t_pov_bool==True)&(' + geo + '==@iloc)'),
['pcwgt', 't_pov_cons']].prod(axis=1).sum(
level=['hazard', 'rp', 'decile']) /
df.loc[df.eval('(t_pov_bool==True)&(' + geo + '==@iloc)'),
'pcwgt'].sum(level=['hazard', 'rp', 'decile']))
df_dec.to_csv('../output_country/' + myC + '/poverty_by_decile.csv')
######################
# Scatter plot of hh that have to delay reconstruction
upper_lim = 1E15
df['t_reco'] = (np.log(1.0 / 0.05) /
df['hh_reco_rate']).clip(upper=upper_lim)
means = []
xmax = 2.5E5
step = xmax / 10.
for i in np.linspace(0, 10, 10):
means.append(df.loc[df.eval(
'(rp==1000)&(c>@i*@step)&(c<=(@i+1)*@step)&(t_reco!=@upper_lim)'
), ['pcwgt', 't_reco']].prod(axis=1).sum() / df.loc[df.eval(
'(rp==1000)&(c>@i*@step)&(c<=(@i+1)*@step)&(t_reco!=@upper_lim)'),
['pcwgt']].sum())
ax = df.loc[df.eval('(c<@xmax)&(t_reco<12)')].plot.hexbin('c',
't_reco',
gridsize=25,
mincnt=1)
plt.plot(step * np.linspace(0, 10, 10), means, zorder=100)
# Do the formatting
ax = title_legend_labels(ax,
'Precipitation flood in ' + myC,
lab_x='Pre-disaster consumption [PhP per cap]',
lab_y='Time to reconstruct [years]',
leg_fs=9)
# Do the saving
plt.draw()
plt.gca().get_figure().savefig('../output_plots/' + myC +
'/t_start_reco_scatter.pdf',
format='pdf')
plt.cla()
######################
# Latex table of poorest quintile poverty time
_cons_to_tex = df_dec.drop([
i for i in df_dec.columns
if i not in ['t_pov_cons_avg_' + j for j in all_geo]
],
axis=1)
_inc_to_tex = df_dec.drop([
i for i in df_dec.columns
if i not in ['t_pov_inc_avg_' + j for j in all_geo]
],
axis=1)
_cons_to_tex = _cons_to_tex.rename(
columns={'t_pov_cons_avg_' + j: j
for j in all_geo}).stack()
_inc_to_tex = _inc_to_tex.rename(
columns={'t_pov_inc_avg_' + j: j
for j in all_geo}).stack()
_cons_to_tex.index.names = ['hazard', 'rp', 'decile', geo]
_inc_to_tex.index.names = ['hazard', 'rp', 'decile', geo]
_to_tex = pd.DataFrame(index=_cons_to_tex.index)
_to_tex['Income'] = _inc_to_tex.round(1)
_to_tex['Consumption'] = _cons_to_tex.round(1)
_to_tex = _to_tex.reset_index().set_index(geo)
_to_tex = _to_tex.loc[_to_tex.eval(
'(hazard=="PF")&(rp==10)&(decile==1)')].sort_values('Consumption',
ascending=False)
print(_to_tex.head())
_to_tex[['Income',
'Consumption']].to_latex('latex/' + myC + '_poverty_duration.tex')
######################
# Plot consumption and income poverty (separately)
df_dec = df_dec.reset_index()
_lab = {
't_pov_cons_avg':
'Average time to exit poverty\n(income net of reconstruction & savings) [months]',
't_pov_inc_avg': 'Average time to exit poverty (income only) [months]'
}
print(df_dec.columns)
for ipov in ['t_pov_cons_avg', 't_pov_inc_avg']:
# Do the plotting
#ax = df_dec.loc[df_dec.eval('(hazard=="PF")&(rp==10)')].plot.scatter('decile',ipov+'no',color=sns_pal[1],lw=0,label='Natl. average (RP = 5 years)',zorder=99)
#df_dec.loc[df_dec.eval('(hazard=="PF")&(rp==1000)')].plot.scatter('decile',ipov+'no',color=sns_pal[3],lw=0,label='Natl. average (RP = 1000 years)',zorder=98,ax=ax)
ax = df_dec.loc[df_dec.eval('(hazard=="PF")&(rp==10)')].plot(
'decile', ipov + 'no', color=sns_pal[1], zorder=97, label='')
df_dec.loc[df_dec.eval('(hazard=="PF")&(rp==1000)')].plot(
'decile',
ipov + 'no',
color=sns_pal[3],
zorder=96,
label='',
ax=ax)
icol = 4
# Which areas to plot?
if myC == 'SL': focus = ['Rathnapura', 'Colombo', 'Kandy', 'Gampaha']
elif myC == 'PH': focus = ['NCR']
for iloc in focus:
df_dec.loc[df_dec.eval('(hazard=="PF")&(rp==10)')].plot.scatter(
'decile',
ipov + '_' + iloc,
color=sns_pal[icol],
lw=0,
label=iloc + ' (RP = 5 years)',
zorder=95,
ax=ax)
df_dec.loc[df_dec.eval('(hazard=="PF")&(rp==10)')].plot(
'decile',
ipov + '_' + iloc,
color=sns_pal[icol],
zorder=94,
label='',
ax=ax)
icol += 1
# Do the formatting
ax = title_legend_labels(ax,
'Precipitation flood in ' + myC,
lab_x='Decile',
lab_y=_lab[ipov],
lim_x=[0.5, 10.5],
lim_y=[-0.1, 42],
leg_fs=9)
ax.xaxis.set_ticks([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
ax.yaxis.set_ticks([0, 6, 12, 18, 24, 30, 36, 42])
# Do the saving
ax.get_figure().savefig('../output_plots/' + myC + '/' + ipov +
'_by_decile.pdf',
format='pdf')
plt.cla()
######################
# Plot consumption and income poverty (separately), with alternative SPs
_lab = {
't_pov_cons_avg':
'Average time to exit poverty\n(income net of reconstruction & savings) [months]',
't_pov_inc_avg': 'Average time to exit poverty (income only) [months]'
}
icol = 0
ax = plt.gca()
for ipov in ['t_pov_cons_avg', 't_pov_inc_avg']:
# Do the plotting
_df_5 = df_dec.loc[df_dec.eval('(hazard=="PF")&(rp==10)')].copy()
_df_1000 = df_dec.loc[df_dec.eval('(hazard=="PF")&(rp==1000)')].copy()
for iSP in _sp:
plt.fill_between(_df_5['decile'].values,
_df_5[ipov + iSP],
_df_1000[ipov + iSP],
alpha=0.3)
_df_5.plot.scatter('decile',
ipov + iSP,
lw=0,
label='',
zorder=99,
ax=ax)
_df_1000.plot.scatter('decile',
ipov + iSP,
lw=0,
label='',
zorder=98,
ax=ax)
_df_5.plot('decile',
ipov + iSP,
zorder=97,
linestyle=':',
label=iSP + ' natl. average (RP = 5 years)',
ax=ax)
_df_1000.plot('decile',
ipov + iSP,
zorder=96,
label=iSP + ' natl. average (RP = 1000 years)',
ax=ax)
icol += 1
# Do the formatting
ax = title_legend_labels(ax,
'Precipitation flood in ' + myC,
lab_x='Decile',
lab_y=_lab[ipov],
lim_x=[0.5, 10.5],
lim_y=[-0.1],
leg_fs=9)
ax.xaxis.set_ticks([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
ax.yaxis.set_ticks([0, 3, 6, 9, 12, 15, 18])
# Do the saving
ax.get_figure().savefig('../output_plots/' + myC + '/' + ipov +
'_with_sps_by_decile.pdf',
format='pdf')
plt.cla()
plt.close('all')
return True
def poverty_gap_plot(
myC,
_myiah,
economy='region_est2',
spvalue=27272.73,
annual_pov_line=81.7 * 365,
annual_pov_line_extreme=41.6 * 365,
output_dir='/Users/jaycocks/Projects/wb_resilience/output_plots/HT/transfer_poverty_'
):
#df = pd.read_csv('/Users/jaycocks/Projects/wb_resilience/intermediate/HT/cat_info.csv')
# _myiah = iah.loc[(iah['hazard']=='EQ')&(iah['affected_cat']=='a')&(iah['helped_cat']=='helped'),]
#_myiah = df
#annual_pov_line = 81.7 * 365
#annual_pov_line_extreme = 41.6 * 365
_myiah['pov_line'] = annual_pov_line
_myiah['sub_line'] = annual_pov_line_extreme
#_myiah['pov_line'] = annual_pov_line_extreme
#_myiah['sub_line'] = annual_pov_line_extreme * 0.7
_myiah['dc0'] = _myiah['pcinc'] #initial
_myiah['dc_pre_reco'] = _myiah['pcinc'] + (spvalue / _myiah['hhsize']
) #final
_myiah = _myiah.reset_index().set_index([economy, 'hhid'])
# use c_initial (dc0) & i_pre_reco (dc_pre_reco)
district_df = _myiah.groupby(economy)['pcwgt'].sum().to_frame(
name='population')
district_df['pop_in_poverty_initial'] = _myiah.loc[
_myiah.ispoor == 1., ].groupby(economy)['pcwgt'].sum()
district_df['pop_in_poverty_final'] = _myiah.loc[
_myiah.dc_pre_reco < annual_pov_line, ].groupby(
economy)['pcwgt'].sum()
district_df['pop_in_extpoverty_initial'] = _myiah.loc[
_myiah.ispoor_extreme == 1., ].groupby(economy)['pcwgt'].sum()
district_df['pop_in_extpoverty_final'] = _myiah.loc[
_myiah.dc_pre_reco < annual_pov_line_extreme, ].groupby(
economy)['pcwgt'].sum()
_myiah['i_initialn'] = _myiah[['pcwgt', 'dc0']].product(axis=1)
district_df['i_initial'] = _myiah.loc[_myiah.ispoor == 1., ].groupby(
economy)['i_initialn'].sum() / district_df['pop_in_poverty_initial']
# district_df['i_final'] = _myiah.loc[_myiah.i_pre_reco<_myiah.pov_line].eval('pcwgt_no*i_pre_reco').sum(level='district')/district_df['pop_in_poverty_final']
_myiah['i_finaln'] = _myiah[['pcwgt', 'dc_pre_reco']].product(axis=1)
district_df['i_final'] = _myiah.loc[_myiah.ispoor == 1., ].groupby(
economy)['i_finaln'].sum() / district_df['pop_in_poverty_initial']
## This is the poverty gap - the average income by region as a percent of the poverty line
# 70 percent is 70 not 0.70 (see the 1E2 multiplier)
district_df['gap_initial'] = 1E2 * _myiah.loc[_myiah.ispoor == 1.].eval(
'pcwgt*(dc0/pov_line)').sum(
level=economy) / district_df['pop_in_poverty_initial']
##district_df['gap_final'] = _myiah.loc[_myiah.i_pre_reco<_myiah.pov_line].eval('pcwgt_no*(i_pre_reco/pov_line)').sum(level=economy)/district_df['pop_in_poverty_final']
district_df['gap_final'] = 1E2 * _myiah.loc[_myiah.ispoor == 1.].eval(
'pcwgt*(dc_pre_reco/pov_line)').sum(
level=economy) / district_df['pop_in_poverty_initial']
## This is the poverty shortfall - the average income by region as a percent of the poverty line
district_df['short_initial'] = _myiah.loc[_myiah.ispoor == 1.].eval(
'-1E2*pcwgt*(1.-dc0/pov_line)').sum(
level=economy) / district_df['pop_in_poverty_initial']
#district_df['gap_final'] = _myiah.loc[_myiah.i_pre_reco<_myiah.pov_line].eval('pcwgt_no*(i_pre_reco/pov_line)').sum(level='district')/district_df['pop_in_poverty_final']
district_df['short_final'] = _myiah.loc[_myiah.ispoor == 1.].eval(
'-1E2*pcwgt*(1.-dc_pre_reco/pov_line)').sum(
level=economy) / district_df['pop_in_poverty_initial']
#
district_df = district_df.sort_values('gap_initial', ascending=False)
#
for _fom in ['gap', 'short']:
plt.close('all')
#
plt.scatter(district_df.index,
district_df[_fom + '_initial'],
alpha=0.6,
color=sns.color_palette('tab20b', n_colors=20)[15],
s=15)
plt.scatter(district_df.index,
district_df[_fom + '_final'],
alpha=0.6,
color=sns.color_palette('tab20b', n_colors=20)[12],
s=15)
#
_xlim = plt.gca().get_xlim()
#
if _fom == 'gap':
_fsub = float(
_myiah.eval('1E2*pcwgt*sub_line').sum() /
_myiah.eval('pcwgt*pov_line').sum())
if _fom == 'short':
_fsub = float(
_myiah.eval('-1E2*pcwgt*(pov_line-sub_line)').sum() /
_myiah.eval('pcwgt*pov_line').sum())
# Subsistence line - the extreme poverty line
plt.plot([_xlim[0], len(district_df.index) - 0.5], [_fsub, _fsub],
color=greys_pal[4],
lw=1.0,
alpha=0.95)
if _fom == 'short':
plt.plot([_xlim[0], len(district_df.index) - 0.5], [0, 0],
color=greys_pal[4],
lw=1.0,
alpha=0.95)
#plt.plot([_xlim[0], len(district_df.index) - 0.5], [_fsub, _fsub], lw=1.0, alpha=0.95)
#if _fom == 'short': plt.plot([_xlim[0], len(district_df.index) - 0.5], [0, 0], lw=1.0,alpha=0.95)
plt.xlim(_xlim)
#
plt.annotate('Poverty gap\npre-transfer',
xy=(len(district_df.index) - 1,
district_df[_fom + '_initial'][-1]),
xytext=(len(district_df.index),
district_df[_fom + '_initial'][-1]),
arrowprops=dict(
arrowstyle="-",
facecolor=greys_pal[5],
connectionstyle="angle,angleA=0,angleB=-90,rad=5"),
fontsize=6.5,
color=greys_pal[7],
va='bottom',
style='italic')
plt.annotate(
'Poverty gap\npost-transfer',
xy=(len(district_df.index) - 1, district_df[_fom + '_final'][-1]),
xytext=(len(district_df.index), district_df[_fom + '_final'][-1]),
arrowprops=dict(arrowstyle="-",
facecolor=greys_pal[5],
connectionstyle="angle,angleA=0,angleB=-90,rad=5"),
fontsize=6.5,
color=greys_pal[7],
va='top',
style='italic')
plt.annotate('Extreme\npoverty\n',
xy=(len(district_df.index) - 1, _fsub),
xytext=(len(district_df.index), _fsub),
arrowprops=dict(
arrowstyle="-",
facecolor=greys_pal[5],
connectionstyle="angle,angleA=0,angleB=-90,rad=5"),
fontsize=6.5,
color=greys_pal[7],
va='bottom',
style='italic')
if _fom == 'short':
plt.annotate(
'Poverty\nline',
xy=(len(district_df.index) - 1, 0),
xytext=(len(district_df.index), 0),
arrowprops=dict(
arrowstyle="-",
facecolor=greys_pal[5],
connectionstyle="angle,angleA=0,angleB=-90,rad=5"),
fontsize=6.5,
color=greys_pal[7],
va='bottom',
style='italic')
for _n, _ in enumerate(district_df.index):
#if _fom == 'gap': plt.plot([_n,_n],[0,district_df.iloc[_n][_fom+'_initial']],color=greys_pal[4],alpha=0.5,ls=':',lw=0.5)
#if _fom == 'short': plt.plot([_n,_n],[0,district_df.iloc[_n][_fom+'_final']],color=greys_pal[4],alpha=0.5,ls=':',lw=0.5)
if _fom == 'gap':
plt.plot([_n, _n],
[0, district_df.iloc[_n][_fom + '_initial']],
alpha=0.5,
ls=':',
lw=0.5)
if _fom == 'short':
plt.plot([_n, _n], [0, district_df.iloc[_n][_fom + '_final']],
alpha=0.5,
ls=':',
lw=0.5)
# Do the formatting
if _fom == 'gap':
_ylabel = 'Average income, as % of poverty line,\nfor population in poverty before Transfer)'
plt.ylim(30, 90)
ax = title_legend_labels(plt.gca(),
'',
lab_x='',
lab_y=_ylabel,
leg_fs=9,
do_leg=False)
ax.set_xticklabels(labels=district_df.index,
rotation=45,
ha='right',
fontsize=8)
sns.despine()
if _fom == 'short':
_ylabel = 'Average income shortfall, as % poverty line,\nfor population in poverty before Transfer)'
ax = title_legend_labels(plt.gca(),
'',
lab_x='',
lab_y=_ylabel,
leg_fs=9,
do_leg=False)
ax.set_xticklabels(labels=district_df.index,
rotation=45,
ha='right',
fontsize=8)
ax.xaxis.set_ticks_position('top')
#ax.invert_yaxis()
#ax.xaxis.tick_top()
plt.ylim(-60, 2)
ax.tick_params(labelbottom='off', labeltop='on')
sns.despine(bottom=True)
plt.grid(False)
# Do the saving
plt.draw()
plt.gca().get_figure().savefig(output_dir + _fom + str(spvalue) +
'.pdf',
format='pdf',
bbox_inches='tight')
plt.cla()
return district_df
def poverty_gap_plot(myC,_myiah,event_level,myHaz,drop_spots=None,_mapres=3000):
# use c_initial & i_pre_reco
district_df = _myiah['pcwgt_no'].sum(level='district').to_frame(name='population')
district_df['pop_in_poverty_initial'] = _myiah.loc[_myiah.ispoor==1.,'pcwgt_no'].sum(level='district')
district_df['pop_in_poverty_final'] = _myiah.loc[_myiah.i_pre_reco<_myiah.pov_line,'pcwgt_no'].sum(level='district')
#
district_df['i_initial'] = _myiah.loc[_myiah.ispoor==1.].eval('pcwgt_no*c_initial').sum(level='district')/district_df['pop_in_poverty_initial']
#district_df['i_final'] = _myiah.loc[_myiah.i_pre_reco<_myiah.pov_line].eval('pcwgt_no*i_pre_reco').sum(level='district')/district_df['pop_in_poverty_final']
district_df['i_final'] = _myiah.loc[_myiah.ispoor==1.].eval('pcwgt_no*i_pre_reco').sum(level='district')/district_df['pop_in_poverty_initial']
#
district_df['gap_initial'] = 1E2*_myiah.loc[_myiah.ispoor==1.].eval('pcwgt_no*(c_initial/pov_line)').sum(level='district')/district_df['pop_in_poverty_initial']
#district_df['gap_final'] = _myiah.loc[_myiah.i_pre_reco<_myiah.pov_line].eval('pcwgt_no*(i_pre_reco/pov_line)').sum(level='district')/district_df['pop_in_poverty_final']
district_df['gap_final'] = 1E2*_myiah.loc[_myiah.ispoor==1.].eval('pcwgt_no*(i_pre_reco/pov_line)').sum(level='district')/district_df['pop_in_poverty_initial']
#
district_df['short_initial'] = _myiah.loc[_myiah.ispoor==1.].eval('-1E2*pcwgt_no*(1.-c_initial/pov_line)').sum(level='district')/district_df['pop_in_poverty_initial']
#district_df['gap_final'] = _myiah.loc[_myiah.i_pre_reco<_myiah.pov_line].eval('pcwgt_no*(i_pre_reco/pov_line)').sum(level='district')/district_df['pop_in_poverty_final']
district_df['short_final'] = _myiah.loc[_myiah.ispoor==1.].eval('-1E2*pcwgt_no*(1.-i_pre_reco/pov_line)').sum(level='district')/district_df['pop_in_poverty_initial']
#
district_df = district_df.sort_values('gap_initial',ascending=False)
#
for _fom in ['gap','short']:
plt.close('all')
#
plt.scatter(district_df.index,district_df[_fom+'_initial'],alpha=0.6,color=sns.color_palette('tab20b', n_colors=20)[15],s=15)
plt.scatter(district_df.index,district_df[_fom+'_final'],alpha=0.6,color=sns.color_palette('tab20b', n_colors=20)[12],s=15)
#
_xlim = plt.gca().get_xlim()
#
if _fom == 'gap': _fsub = float(_myiah.eval('1E2*pcwgt_no*sub_line').sum()/_myiah.eval('pcwgt_no*pov_line').sum())
if _fom == 'short': _fsub = float(_myiah.eval('-1E2*pcwgt_no*(pov_line-sub_line)').sum()/_myiah.eval('pcwgt_no*pov_line').sum())
# Subsistence line
plt.plot([_xlim[0],len(district_df.index)-0.5],[_fsub,_fsub],color=greys_pal[4],lw=1.0,alpha=0.95)
if _fom == 'short': plt.plot([_xlim[0],len(district_df.index)-0.5],[0,0],color=greys_pal[4],lw=1.0,alpha=0.95)
plt.xlim(_xlim)
#
plt.annotate('Poverty gap\npre-Idai',xy=(len(district_df.index)-1,district_df[_fom+'_initial'][-1]),xytext=(len(district_df.index),district_df[_fom+'_initial'][-1]),
arrowprops=dict(arrowstyle="-",facecolor=greys_pal[5],connectionstyle="angle,angleA=0,angleB=-90,rad=5"),fontsize=6.5,color=greys_pal[7],va='bottom',style='italic')
plt.annotate('Poverty gap\npost-Idai',xy=(len(district_df.index)-1,district_df[_fom+'_final'][-1]),xytext=(len(district_df.index),district_df[_fom+'_final'][-1]),
arrowprops=dict(arrowstyle="-",facecolor=greys_pal[5],connectionstyle="angle,angleA=0,angleB=-90,rad=5"),fontsize=6.5,color=greys_pal[7],va='top',style='italic')
plt.annotate('Extreme\npoverty\n',xy=(len(district_df.index)-1,_fsub),xytext=(len(district_df.index),_fsub),
arrowprops=dict(arrowstyle="-",facecolor=greys_pal[5],connectionstyle="angle,angleA=0,angleB=-90,rad=5"),fontsize=6.5,color=greys_pal[7],va='bottom',style='italic')
if _fom == 'short': plt.annotate('Poverty\nline',xy=(len(district_df.index)-1,0),xytext=(len(district_df.index),0),
arrowprops=dict(arrowstyle="-",facecolor=greys_pal[5],connectionstyle="angle,angleA=0,angleB=-90,rad=5"),
fontsize=6.5,color=greys_pal[7],va='bottom',style='italic')
#
for _n,_ in enumerate(district_df.index):
if _fom == 'gap': plt.plot([_n,_n],[0,district_df.iloc[_n][_fom+'_initial']],color=greys_pal[4],alpha=0.5,ls=':',lw=0.5)
if _fom == 'short': plt.plot([_n,_n],[0,district_df.iloc[_n][_fom+'_final']],color=greys_pal[4],alpha=0.5,ls=':',lw=0.5)
#
# Do the formatting
if _fom == 'gap':
_ylabel = 'Poverty gap (Average income, as % of poverty line,\nfor households in poverty before Idai)'
plt.ylim(45,85)
ax = title_legend_labels(plt.gca(),'',lab_x='',lab_y=_ylabel,leg_fs=9,do_leg=False)
ax.set_xticklabels(labels=district_df.index,rotation=45,ha='right',fontsize=8)
sns.despine()
if _fom == 'short':
_ylabel = 'Poverty gap (Average income shortfall, as % poverty line,\nfor households in poverty before Idai)'
ax = title_legend_labels(plt.gca(),'',lab_x='',lab_y=_ylabel,leg_fs=9,do_leg=False)
ax.set_xticklabels(labels=district_df.index,rotation=45,ha='right',fontsize=8)
ax.xaxis.set_ticks_position('top')
#ax.invert_yaxis()
#ax.xaxis.tick_top()
plt.ylim(-50,2)
ax.tick_params(labelbottom='off',labeltop='on')
sns.despine(bottom=True)
plt.grid(False)
# Do the saving
plt.draw()
plt.gca().get_figure().savefig('../output_plots/'+myC+'/idai_poverty_'+_fom+'.pdf',format='pdf',bbox_inches='tight')
plt.cla()
def draw_drought_income_plot(df=None, _='initial', _rho=0.70):
big_pal = sns.color_palette('tab20b', n_colors=20)
if df is None:
df = pd.read_csv('../output_country/MW/out_ag.csv').set_index(
event_level + ['hhid'])
df['frac_ag'] = df.eval('(di_ag/v_ag)/c')
df['i_nonag'] = 1E-3 * df.eval('c*(1-frac_ag)')
df['i_ag'] = 1E-3 * df.eval('c*(frac_ag)')
df = (df.loc[(df.frac_ag >= 0.70) & (df.frac_ag < 0.99) &
(df.v_ag <= 0.40)].sample(1)).squeeze()
print(df)
## Non-agricultural income
#plt.plot([0,2],[df['c_nonag'],df['c_nonag']])
#########################
# Agricultural income spenddown
_x = []
_y = []
_x_lean = []
_y_lean = []
_xDR = []
_yDR = []
_xDR_lean = []
_yDR_lean = []
t_series, _dt = np.linspace(0, 2, 4E2, retstep=True)
c_ag_init = _rho * df['i_ag'] / (1 - np.exp(-_rho))
for _t in t_series:
_sf = 1 if _t < 1 else (1 - df['v_ag'])
_tcyc = _t if _t < 1 else _t - 1
#
_cag_t = c_ag_init * _sf * np.exp(-_rho * _tcyc)
#
if _t <= 1:
_x.append(_t)
_y.append(_cag_t)
if _cag_t <= df['i_ag'] * _sf:
_x_lean.append(_t)
_y_lean.append(_cag_t)
else:
_xDR.append(_t)
_yDR.append(_cag_t)
if _cag_t <= df['i_ag'] * _sf:
_xDR_lean.append(_t)
_yDR_lean.append(_cag_t)
# Plot c_ag(t)
plt.plot(_x, _y, color=greys_pal[7], alpha=0.9, zorder=100)
plt.plot(_xDR, _yDR, color=greys_pal[7], alpha=0.9, zorder=100)
#
plt.plot([_ - 1 for _ in _x],
_y,
color=greys_pal[7],
alpha=0.9,
zorder=100,
clip_on=True)
plt.plot([_ + 2 for _ in _x],
_y,
color=greys_pal[7],
alpha=0.9,
zorder=100,
clip_on=True)
#
plt.plot([0, 0], [_y[-1], _y[0]],
color=greys_pal[7],
alpha=0.9,
zorder=100) # connect year = -1 to year = 0
plt.plot([1, 1], [_y[-1], _yDR[0]],
color=greys_pal[8],
alpha=0.9,
zorder=100) # connect year = 0 to year = 1
plt.plot([2, 2], [_yDR[-1], _y[0]],
color=greys_pal[7],
alpha=0.9,
zorder=100) # connect year = 1 to year = 2
# and do annotation
#_str = 'Household consumption\nof agricultural income\n$c_{ag}(t)\,=\,C^{\prime}_{ag}\cdot e^{-\gamma_{ag}t}$\n(cf. Eqs. 3$\,$&$\,$7)'
_str = 'Household consumption\nof agricultural income $c_{ag}(t)$\n(cf.$\,$Eqs.$\,$2-7)'
plt.annotate(_str,
xy=(_x[5], _y[1]),
xytext=(_x[13], _y[0]),
arrowprops=dict(
arrowstyle="-",
facecolor=greys_pal[7],
connectionstyle="angle,angleA=0,angleB=-135,rad=0"),
fontsize=7.5,
color=greys_pal[7],
va='bottom',
ha='left',
linespacing=1.4)
#########################
# Agricultural income spike
for yr_no in [0, 1, 2]:
# annual income squeezed into 1 month
len_harvest = 4 # months
_sf = 1
if yr_no == 1: _sf = (1 - df['v_ag'])
plt.plot([yr_no, yr_no], [0, (12 / len_harvest) * df['i_ag'] * _sf],
color=big_pal[10],
alpha=0.5)
plt.plot([yr_no, yr_no + len_harvest / 12],
[(12 / len_harvest) * df['i_ag'] * _sf,
(12 / len_harvest) * df['i_ag'] * _sf],
color=big_pal[10],
alpha=0.5)
plt.fill_between([yr_no, yr_no + len_harvest / 12],
[(12 / len_harvest) * df['i_ag'] * _sf,
(12 / len_harvest) * df['i_ag'] * _sf],
color=big_pal[11],
alpha=0.5)
plt.plot([yr_no + len_harvest / 12, yr_no + len_harvest / 12],
[0, (12 / len_harvest) * df['i_ag'] * _sf],
color=big_pal[10],
alpha=0.5)
#
# Mark lean season
if yr_no != 2:
plt.fill_between(
(_x_lean if yr_no == 0 else _xDR_lean)[:-1],
y1=(_y_lean if yr_no == 0 else _yDR_lean)[:-1],
#y2=[0 for _tcyc in _y_lean[:-1]],
color=big_pal[13],
alpha=0.1)
# and do annotation
if yr_no == 0:
plt.annotate('Lean season',
xy=((_xDR_lean if yr_no == 1 else _x_lean)[-4],
df['i_ag'] * (0.03)),
fontsize=7,
color=big_pal[13],
va='bottom',
ha='right')
#
# Plot annualized income as dotted line
plt.plot([yr_no, yr_no + 1], [df['i_ag'] * _sf, df['i_ag'] * _sf],
color=greys_pal[4],
alpha=0.4,
ls='--',
lw=0.75)
# and do annotation
_labstr = ('Income from main\nharvest, normal year' if yr_no == 0
else 'Income from main\nharvest, drought year')
plt.annotate(
_labstr,
xy=(yr_no + len_harvest / 12 * 0.20,
(12 / len_harvest) * df['i_ag'] * _sf),
xytext=(yr_no + len_harvest / 12 * (0.20 * 1.75),
(12 / len_harvest) * df['i_ag'] * (_sf + 0.06)),
arrowprops=dict(
arrowstyle="-",
facecolor=greys_pal[7],
connectionstyle="angle,angleA=0,angleB=-90,rad=5"),
fontsize=7.5,
color=greys_pal[7],
va='center',
ha='left')
# ticks
plt.xticks([0, 0.25, 0.5, 0.75, 1.0, 1.25, 1.50, 1.75, 2])
plt.gca().set_xticklabels(labels=[
'April', 'July', 'October', 'January', 'April', 'July', 'October',
'January', 'April'
],
ha='right',
fontsize=8,
rotation=45)
plt.tick_params(axis='x', which='major', color=greys_pal[4])
plt.tick_params(axis='y', which='major', color=greys_pal[4], labelsize=8)
#
title_legend_labels(plt.gca(),
'',
lab_x='',
lab_y='Kwacha per capita (,000)',
lim_x=None,
lim_y=None,
leg_fs=9,
do_leg=False)
#
plt.grid(False)
plt.xlim(-0.05, 2.05)
plt.ylim(0)
sns.despine()
plt.gcf().savefig('../output_plots/MW/drought/schematic.pdf',
format='pdf',
bbox_inches='tight')
plt.close('all')
def plot(self, c1=None, c2=None, c3=None, c4=None):
x_shift = 2.
# Plot the fraction of winners
plt.plot([0, 100], [0, 0], linewidth=1.5, color=greys[5], zorder=1)
for _c in [c1, c2, c3, c4]:
if _c is not None:
_ls = '-'
if _c.incl_targeting_corr is not None: _ls = '--'
plt.plot(_c.scale_out,
_c.impact_winners,
color=_c.pdict['scaleout_' + _c.sp + '_100col'],
linestyle=_ls)
#####
# Find/plot/annotate max
#plt.scatter(_c.scale_out[_c.impact_winners.index(max(_c.impact_winners))],max(_c.impact_winners),color=_c.pdict['scaleout_'+_c.sp+'_100col'])
_opty = max(_c.impact_winners)
_optx = _c.scale_out[_c.impact_winners.index(_opty)]
plt.plot([_optx, _optx], [_opty, _opty + 0.5],
color=greys[4],
linewidth=1.2)
_init_enroll = 100. * (_c.fracs[1] + _c.fracs[2]) / 2.
plt.annotate('Increase by\n' +
str(round(_optx - _init_enroll, 1)) +
'% in all quintiles',
xy=(_optx, _opty + 0.75),
size=5,
va='bottom',
ha='center')
#####
if _c.incl_targeting_corr is None:
_anno = _c.sp.replace('ct', 'Cash\ntransfer').replace(
'trsgob', 'Non-cash\ntransfer')
plt.annotate(_anno,
xy=(_c.scale_out[0] - x_shift,
_c.impact_winners[0]),
weight='bold',
color=greys[6],
ha='right',
va='center',
size=5.5)
else:
_anno = 'Reduce inclusion\nerror by 10%'
plt.annotate(_anno,
xy=(_c.scale_out[0] + x_shift,
_c.impact_winners[0]),
weight='bold',
color=greys[5],
ha='left',
va='center',
size=5)
plt.xlim(0, 100)
lgd = title_legend_labels(
plt.gca(),
iso_to_name[self.pais],
lab_x='Scale out (% of each quintile enrolled)',
lab_y='Percent of poorest 40% that benefits',
do_leg=False)
sns.despine(bottom=True)
plt.gcf().savefig('output/sp/' + self.pais + '_optimal_sp_winners.pdf',
format='pdf',
bbox_inches='tight')
plt.cla()
# Plot relative impact (summed at quintile-level)
plt.plot([0, 100], [0, 0], linewidth=1.5, color=greys[5], zorder=1)
for _c in [c1, c2, c3, c4]:
if _c is not None:
_ls = '-'
if _c.incl_targeting_corr is not None: _ls = '--'
plt.plot(_c.scale_out,
_c.impact_rel,
color=_c.pdict['scaleout_' + _c.sp + '_100col'],
linestyle=_ls)
if _c.incl_targeting_corr is None:
_anno = _c.sp.replace('ct', 'Cash\ntransfer').replace(
'trsgob', 'Non-cash\ntransfer')
plt.annotate(_anno,
xy=(_c.scale_out[0] - x_shift,
_c.impact_rel[0]),
weight='bold',
color=greys[6],
ha='right',
va='center',
size=5.5)
else:
_anno = 'Reduce inclusion\nerror by ' + str(
int(100 * _c.incl_targeting_corr)) + '%'
plt.annotate(_anno,
xy=(_c.scale_out[0] + x_shift,
_c.impact_rel[0]),
weight='bold',
color=greys[5],
ha='left',
va='center',
size=5)
plt.xlim(0, 100)
lgd = title_legend_labels(
plt.gca(),
iso_to_name[self.pais],
lab_x='Scale out (% of bottom 40% enrolled)',
lab_y=
'Net impact on bottom 40%\n(relative to quintile expenditures)',
do_leg=False)
sns.despine(bottom=True)
plt.gcf().savefig('output/sp/' + self.pais +
'_optimal_sp_rel_impact.pdf',
format='pdf',
bbox_inches='tight')
plt.cla()
# Plot relative impact (calculated at household level)
plt.plot([0, 100], [0, 0], linewidth=1.5, color=greys[5], zorder=1)
for _c in [c1, c2, c3, c4]:
if _c is not None:
_ls = '-'
if _c.incl_targeting_corr is not None: _ls = '--'
plt.plot(_c.scale_out,
_c.impact_rel_by_hh,
color=_c.pdict['scaleout_' + _c.sp + '_100col'],
linestyle=_ls)
if _c.incl_targeting_corr is None:
_anno = _c.sp.replace('ct', 'Cash\ntransfer').replace(
'trsgob', 'Non-cash\ntransfer')
plt.annotate(_anno,
xy=(_c.scale_out[0] - x_shift,
_c.impact_rel_by_hh[0]),
weight='bold',
color=greys[6],
ha='right',
va='center',
size=5.5)
else:
_anno = 'Reduce inclusion\nerror by ' + str(
int(100 * _c.incl_targeting_corr)) + '%'
plt.annotate(_anno,
xy=(_c.scale_out[0] + x_shift,
_c.impact_rel_by_hh[0]),
weight='bold',
color=greys[4],
ha='left',
va='center',
size=5)
plt.xlim(0, 100)
lgd = title_legend_labels(
plt.gca(),
iso_to_name[self.pais],
lab_x='Scale out (% of bottom 40% enrolled)',
lab_y=
'Net impact on bottom 40%\n(relative to household expenditures)',
do_leg=False)
sns.despine(bottom=True)
plt.gcf().savefig('output/sp/' + self.pais +
'_optimal_sp_rel_impact_by_hh.pdf',
format='pdf',
bbox_inches='tight')
plt.cla()
# Plot average impact, in $
plt.plot([0, 100], [0, 0], linewidth=1.5, color=greys[5], zorder=1)
for _c in [c1, c2, c3, c4]:
if _c is not None:
_ls = '-'
if _c.incl_targeting_corr is not None: _ls = '--'
plt.plot(_c.scale_out,
_c.impact_abs,
color=_c.pdict['scaleout_' + _c.sp + '_100col'],
linestyle=_ls)
if _c.incl_targeting_corr is None:
_anno = _c.sp.replace('ct', 'Cash\ntransfer').replace(
'trsgob', 'Non-cash\ntransfer')
plt.annotate(_anno,
xy=(_c.scale_out[0] - x_shift,
_c.impact_abs[0]),
weight='bold',
color=greys[6],
ha='right',
va='center',
size=5.5)
else:
_anno = 'Reduce inclusion\nerror by ' + str(
int(100 * _c.incl_targeting_corr)) + '%'
plt.annotate(_anno,
xy=(_c.scale_out[0] + x_shift,
_c.impact_abs[0]),
weight='bold',
color=greys[4],
ha='left',
va='center',
size=5)
plt.xlim(0, 100)
lgd = title_legend_labels(plt.gca(),
iso_to_name[self.pais],
lab_x='Scale out (% of bottom 40% enrolled)',
lab_y='Impact on bottom 40% [INT$]',
do_leg=False)
sns.despine(bottom=True)
plt.gcf().savefig('output/sp/' + self.pais +
'_optimal_sp_abs_impact.pdf',
format='pdf',
bbox_inches='tight')
plt.cla()
plt.close('all')
