def get_mc_directories(rfrac):
rootdir = os.getcwd()+'/monte_carlo/'
_dirs = []; _lbls = []
#
for subdir, dirs, files in os.walk(rootdir):
if subdir != rootdir and rfrac in subdir:
_dirs.append(subdir)
__sector = subdir.replace(rootdir+'relax_','').replace(rfrac,'')
try: _lbls.append(monte_carlo(0,'base').sector_labels[__sector])
except: _lbls.append(__sector)
return _dirs,_lbls
def plot_status_vs_valueadd(base_code,rfrac,fom_x,fom_y):
mc = monte_carlo(0,'base')
dirs,labels = get_mc_directories(rfrac)
base_x = pd.read_csv('monte_carlo/{}/{}.csv'.format(base_code,fom_x),index_col=0).sum(axis=1)
base_y = pd.read_csv('monte_carlo/{}/{}.csv'.format(base_code,fom_y),index_col=0).sum(axis=1)
sf_x = 1
sf_y = 1
for _nd,_d in enumerate(dirs):
_fx = pd.read_csv(_d+'/{}.csv'.format(fom_x),index_col=0)
_fy = pd.read_csv(_d+'/{}.csv'.format(fom_y),index_col=0)
x_val = sf_x*(base_x.mean()-_fx.sum(axis=1).mean())
y_val = sf_y*(base_y.mean()-_fy.sum(axis=1).mean())
x_err = 2*(base_x-_fx.sum(axis=1)).std()/np.sqrt(float(base_x.shape[0]))
y_err = 2*(base_y-_fy.sum(axis=1)).std()/np.sqrt(float(base_y.shape[0]))
if (x_val-x_err > 0 or y_val-y_err > 0):
plt.errorbar(x_val,y_val,xerr=x_err,yerr=y_err,fmt='o',zorder=90,color=sns_pal[1],alpha=(0.7 if (x_val-x_err>0 and y_val-y_err>0) else 0.3))
plt.annotate(labels[_nd],xy=(x_val+x_err/10,y_val+y_err/10),fontsize=8,color=greys_pal[5],rotation=0,ha='left',va='bottom',weight=500)
plt.xlim(0)
plt.ylim(0)
plt.annotate('{}% recovery of\nsectoral activity'.format(rfrac),xy=(0.05,0.95),
xycoords='axes fraction',ha='left',va='top',fontsize=8,color=greys_pal[6],annotation_clip=False)
plt.grid(False)
sns.despine(bottom=True,left=True)
plt.plot([0,0],plt.gca().get_ylim(),color=greys_pal[6],zorder=10)
plt.plot(plt.gca().get_xlim(),[0,0],color=greys_pal[6],zorder=10)
plt.xlabel(ax_label_dict[fom_x],labelpad=10,linespacing=1.75)
plt.ylabel(ax_label_dict[fom_y],labelpad=10,linespacing=1.75)
# save & close
plt.savefig('figs/scatter_{}_{}.pdf'.format(fom_x,fom_y),format='pdf',bbox_inches='tight')
plt.close('all')
def explore_poverty_mysteries(hh_df, scaleup_CCT=0, mirror_subsistence=False):
mc = monte_carlo(0, 'base')
ul = 250
nbins = int(50)
# total income <-- use to structure plot
tot_hgt, _bins = np.histogram(hh_df.eval('pcinc_initial').clip(upper=ul),
bins=nbins,
weights=hh_df['popwgt'])
wid = (_bins[1] - _bins[0])
bin_slice = '(pcinc_initial>@b)&(pcinc_initial<=@_bins[@n+1])'
# households falling into poverty
newpov = '&(@mc.m2d*pcinc_initial>3.20)&(@mc.m2d*pcinc_final<=3.20)'
month1 = '&(@mc.m2d*(hhinc-income_loss+savings/1+cct4P*@scaleup_CCT)/hhsize<=3.20)'
month2 = '&(@mc.m2d*(hhinc-income_loss+savings/1+cct4P*@scaleup_CCT)/hhsize>3.20)&(@mc.m2d*(hhinc-income_loss+savings/2+cct4P*@scaleup_CCT)/hhsize<=3.20)'
month3 = '&(@mc.m2d*(hhinc-income_loss+savings/2+cct4P*@scaleup_CCT)/hhsize>3.20)&(@mc.m2d*(hhinc-income_loss+savings/3+cct4P*@scaleup_CCT)/hhsize<=3.20)'
month4 = '&(@mc.m2d*(hhinc-income_loss+savings/3+cct4P*@scaleup_CCT)/hhsize>3.20)'
# households falling into extreme poverty
newsub = '&(@mc.m2d*pcinc_initial>1.90)&(@mc.m2d*pcinc_final<=1.90)'
month1s = '&(@mc.m2d*(hhinc-income_loss+savings/1+cct4P*@scaleup_CCT)/hhsize<=1.90)'
month2s = '&(@mc.m2d*(hhinc-income_loss+savings/1+cct4P*@scaleup_CCT)/hhsize>1.90)&(@mc.m2d*(hhinc-income_loss+savings/2+cct4P*@scaleup_CCT)/hhsize<=1.90)'
month3s = '&(@mc.m2d*(hhinc-income_loss+savings/2+cct4P*@scaleup_CCT)/hhsize>1.90)&(@mc.m2d*(hhinc-income_loss+savings/3+cct4P*@scaleup_CCT)/hhsize<=1.90)'
month4s = '&(@mc.m2d*(hhinc-income_loss+savings/3+cct4P*@scaleup_CCT)/hhsize>1.90)'
newpov_hgt = []
newpov_1m_hgt = []
newpov_2m_hgt = []
newpov_3m_hgt = []
newpov_4m_hgt = []
newsub_hgt = []
newsub_1m_hgt = []
newsub_2m_hgt = []
newsub_3m_hgt = []
newsub_4m_hgt = []
for n, b in enumerate(_bins[:-1]):
newpov_hgt.append(hh_df.loc[hh_df.eval(bin_slice + newpov),
'popwgt'].sum())
newpov_1m_hgt.append(hh_df.loc[hh_df.eval(bin_slice + newpov + month1),
'popwgt'].sum())
newpov_2m_hgt.append(hh_df.loc[hh_df.eval(bin_slice + newpov + month2),
'popwgt'].sum())
newpov_3m_hgt.append(hh_df.loc[hh_df.eval(bin_slice + newpov + month3),
'popwgt'].sum())
newpov_4m_hgt.append(hh_df.loc[hh_df.eval(bin_slice + newpov + month4),
'popwgt'].sum())
#
newsub_1m_hgt.append(
-1 * hh_df.loc[hh_df.eval(bin_slice + newsub + month1s),
'popwgt'].sum())
newsub_2m_hgt.append(
-1 * hh_df.loc[hh_df.eval(bin_slice + newsub + month2s),
'popwgt'].sum())
newsub_3m_hgt.append(
-1 * hh_df.loc[hh_df.eval(bin_slice + newsub + month3s),
'popwgt'].sum())
newsub_4m_hgt.append(
-1 * hh_df.loc[hh_df.eval(bin_slice + newsub + month4s),
'popwgt'].sum())
#
# plot them
#ax = plt.step(_bins[:-1]+wid,newpov_hgt,linewidth=0.5,alpha=0.4)
btm = 0
lbl = '< 1 month ({} m.)'.format(round(np.sum(newpov_1m_hgt), 1))
ax = plt.bar(_bins[:-1],
newpov_1m_hgt,
bottom=btm,
width=wid,
align='edge',
linewidth=0,
alpha=0.5,
facecolor=reds_pal[0],
label=lbl)
btm = newpov_1m_hgt
lbl = '1-2 months ({} m.)'.format(round(np.sum(newpov_2m_hgt), 1))
ax = plt.bar(_bins[:-1],
newpov_2m_hgt,
width=wid,
bottom=btm,
align='edge',
linewidth=0,
alpha=0.5,
facecolor=reds_pal[1],
label=lbl)
btm = [i + j for i, j in zip(newpov_1m_hgt, newpov_2m_hgt)]
lbl = '2-3 months ({} m.)'.format(round(np.sum(newpov_3m_hgt), 1))
ax = plt.bar(_bins[:-1],
newpov_3m_hgt,
width=wid,
bottom=btm,
align='edge',
linewidth=0,
alpha=0.5,
facecolor=reds_pal[2],
label=lbl)
btm = [
i + j + k
for i, j, k in zip(newpov_1m_hgt, newpov_2m_hgt, newpov_3m_hgt)
]
lbl = '> 3 months ({} m.)'.format(round(np.sum(newpov_4m_hgt), 1))
ax = plt.bar(_bins[:-1],
newpov_4m_hgt,
width=wid,
bottom=btm,
align='edge',
linewidth=0,
alpha=0.5,
facecolor=reds_pal[3],
label=lbl)
# plot hh falling into subsistence
if mirror_subsistence:
btm = 0
lbl = '< 1 month ({} m.)'.format(round(-1 * np.sum(newsub_1m_hgt), 1))
ax = plt.bar(_bins[:-1],
newsub_1m_hgt,
bottom=btm,
width=wid,
align='edge',
linewidth=0,
alpha=0.35,
facecolor=reds_pal[0],
label=lbl)
btm = newsub_1m_hgt
lbl = '1-2 months ({} m.)'.format(-1 * round(np.sum(newsub_2m_hgt), 1))
ax = plt.bar(_bins[:-1],
newsub_2m_hgt,
width=wid,
bottom=btm,
align='edge',
linewidth=0,
alpha=0.35,
facecolor=reds_pal[1],
label=lbl)
btm = [i + j for i, j in zip(newsub_1m_hgt, newsub_2m_hgt)]
lbl = '2-3 months ({} m.)'.format(-1 * round(np.sum(newsub_3m_hgt), 1))
ax = plt.bar(_bins[:-1],
newsub_3m_hgt,
width=wid,
bottom=btm,
align='edge',
linewidth=0,
alpha=0.35,
facecolor=reds_pal[2],
label=lbl)
btm = [
i + j + k
for i, j, k in zip(newsub_1m_hgt, newsub_2m_hgt, newsub_3m_hgt)
]
lbl = '> 3 months ({} m.)'.format(-1 * round(np.sum(newsub_4m_hgt), 1))
ax = plt.bar(_bins[:-1],
newsub_4m_hgt,
width=wid,
bottom=btm,
align='edge',
linewidth=0,
alpha=0.35,
facecolor=reds_pal[3],
label=lbl)
#
plt.legend(title='Shock duration{}'.format('\n({}% CCT scaleup)'.format(
int(1E2 * scaleup_CCT)) if scaleup_CCT != 0 else ''),
labelspacing=0.75,
ncol=1,
fontsize=8,
borderpad=0.75,
fancybox=True,
frameon=True,
framealpha=0.9)
plt.xlabel('pre-COVID income [PPP$/cap/month]', labelpad=10)
plt.ylabel('Impoverished population [millions]', labelpad=10)
if not mirror_subsistence: plt.xlim(90, ul)
else: plt.xlim(0, ul)
plt.grid(False)
sns.despine(left=True, bottom=True)
plt.plot([0, ul], [0, 0], color=greys_pal[4], lw=1)
plt.savefig('figs/new_poverty{}{}.pdf'.format(
'_net_sub' if mirror_subsistence else '',
'_CCT{}x'.format(scaleup_CCT) if scaleup_CCT != 0 else ''),
format='pdf',
bbox_inches='tight')
plt.close('all')
def plot_income_scatter(hh_df, _ul=20, with_migration=True):
mc = monte_carlo(0, 'base')
nbins = 11
dy = 11
plt.scatter(mc.m2d * hh_df.loc[hh_df.income_loss != 0, 'pcinc_initial'],
mc.m2d * hh_df.loc[hh_df.income_loss != 0, 'pcinc_final'],
s=6,
alpha=0.4)
base_hgt, _bins = np.histogram(
(mc.m2d * hh_df['pcinc_initial']).clip(upper=_ul),
bins=[2 * n for n in range(0, 11)],
weights=hh_df['popwgt'])
shock_hgt, _ = np.histogram(
(mc.m2d * hh_df['pcinc_final']).clip(upper=_ul),
bins=_bins,
weights=hh_df['popwgt'])
wid = (_bins[1] - _bins[0])
for n, b in enumerate(_bins):
plt.plot([b, b], [0, _ul],
lw=0.6,
color=greys_pal[5],
ls=':',
zorder=100)
try:
plt.annotate(r'$\Delta$P' + ': {}%'.format(
int(round(1E2 *
(shock_hgt[n] - base_hgt[n]) / base_hgt[n], 0))),
xy=(b + wid / 2, np.e**(np.log(dy) / _ul *
(b + wid / 2)) + _ul / 2 + 1),
color=greys_pal[7],
fontsize=5.5,
ha='center',
va='bottom',
annotation_clip=False)
except:
pass
try:
bin_slice = '(@mc.m2d*pcinc_initial>@b)&(@mc.m2d*pcinc_initial<=@_bins[@n+1])'
i_i = mc.m2d * hh_df.loc[
hh_df.eval(bin_slice), ['popwgt', 'pcinc_initial']].prod(
axis=1).sum() / hh_df.loc[hh_df.eval(bin_slice),
'popwgt'].sum()
if with_migration:
bin_slice = '(@mc.m2d*pcinc_final>@b)&(@mc.m2d*pcinc_final<=@_bins[@n+1])'
else:
bin_slice = '(@mc.m2d*pcinc_initial>@b)&(@mc.m2d*pcinc_initial<=@_bins[@n+1])'
i_f = mc.m2d * hh_df.loc[
hh_df.eval(bin_slice), ['popwgt', 'pcinc_final']].prod(
axis=1).sum() / hh_df.loc[hh_df.eval(bin_slice),
'popwgt'].sum()
di = round(1E2 * (i_f - i_i) / i_i, 1)
plt.annotate(r'$\Delta$i' + ': {}%'.format(di),
xy=(b + wid / 2,
np.e**(np.log(dy) / _ul *
(b + wid / 2)) + _ul / 2 + 0.2),
color=greys_pal[7],
fontsize=5.5,
ha='center',
va='bottom',
annotation_clip=False)
if n < 5:
if n == 0:
plt.annotate('poverty gap',
style='italic',
xy=(b + wid / 2,
np.e**(np.log(dy) / _ul *
(b + wid / 2)) + _ul / 2 - 1),
color=greys_pal[7],
fontsize=5.5,
ha='center',
va='bottom')
pgap_i = round(1E2 * (i_i - 3.2) / 3.2, 1)
plt.annotate(r'initial:' + '\n{}%'.format(pgap_i),
xy=(b + wid / 20,
np.e**(np.log(dy) / _ul *
(b + wid / 2)) + _ul / 2 - 2.2),
color=greys_pal[7],
fontsize=5.5,
ha='left',
va='bottom',
annotation_clip=False)
pgap_f = round(1E2 * (i_f - 3.2) / 3.2, 1)
plt.annotate(r'final:' + '\n{}%'.format(pgap_f),
xy=(b + wid * 19 / 20,
np.e**(np.log(dy) / _ul *
(b + wid / 2)) + _ul / 2 - 3.4),
color=greys_pal[7],
fontsize=5.5,
ha='right',
va='bottom',
annotation_clip=False)
#pgap_i = None
except:
pass
plt.xlabel('Income [PPP$/cap/day]', labelpad=10)
plt.ylabel('Income during shock [PPP$/cap/day]', labelpad=10)
plt.xlim(0, _ul)
plt.ylim(0, _ul)
plt.xticks([2 * n for n in range(0, 11)])
sns.despine()
plt.savefig('figs/income_scatter_{}.pdf'.format(
'with_bin_migrants' if with_migration else 'no_bin_migrants'),
format='pdf',
bbox_inches='tight')
plt.close('all')
def build_shock(fom='totpop_pov'):
pal = monte_carlo().ichan_cols
# create plot showing how impact channels add to affected, poverty
initial = pd.read_csv('monte_carlo/inital_pop_by_class.csv'.format(fom))
base = pd.read_csv('monte_carlo/base/{}.csv'.format(fom),
index_col=0)[['vul', 'sec', 'mc']].sum(axis=1)
W1E0R0 = pd.read_csv('monte_carlo/W1E0R0/{}.csv'.format(fom),
index_col=0)[['vul', 'sec', 'mc']].sum(axis=1)
W0E1R0 = pd.read_csv('monte_carlo/W0E1R0/{}.csv'.format(fom),
index_col=0)[['vul', 'sec', 'mc']].sum(axis=1)
W0E0R1 = pd.read_csv('monte_carlo/W0E0R1/{}.csv'.format(fom),
index_col=0)[['vul', 'sec', 'mc']].sum(axis=1)
W1E1R0 = pd.read_csv('monte_carlo/W1E1R0/{}.csv'.format(fom),
index_col=0)[['vul', 'sec', 'mc']].sum(axis=1)
W0E1R1 = pd.read_csv('monte_carlo/W0E1R1/{}.csv'.format(fom),
index_col=0)[['vul', 'sec', 'mc']].sum(axis=1)
W1E0R1 = pd.read_csv('monte_carlo/W1E0R1/{}.csv'.format(fom),
index_col=0)[['vul', 'sec', 'mc']].sum(axis=1)
wid = 0.8
alp = 0.6
lw = 0.0
plt.barh(7,
W0E0R1.mean(),
facecolor=pal['remits'],
lw=lw,
height=wid,
alpha=alp)
plt.annotate(' {} mil.'.format(round(W0E0R1.mean(), 1)),
xy=(W0E0R1.mean(), 7 + wid / 2),
va='center',
ha='left',
color=greys_pal[6])
plt.barh(6,
W0E1R0.mean(),
facecolor=pal['entrep'],
lw=lw,
height=wid,
alpha=alp)
plt.annotate(' {} mil.'.format(round(W0E1R0.mean(), 1)),
xy=(W0E1R0.mean(), 6 + wid / 2),
va='center',
ha='left',
color=greys_pal[6])
plt.barh(5,
W1E0R0.mean(),
facecolor=pal['nonag_wage'],
lw=lw,
height=wid,
alpha=alp)
plt.annotate(' {} mil.'.format(round(W1E0R0.mean(), 1)),
xy=(W1E0R0.mean(), 5 + wid / 2),
va='center',
ha='left',
color=greys_pal[6])
#
plt.barh(3,
W0E1R1.mean(),
facecolor=pal['entrep'],
lw=lw,
height=wid * 2 / 3,
alpha=alp)
plt.barh(3 + wid / 3,
W0E1R1.mean(),
facecolor=pal['remits'],
lw=lw,
height=wid * 2 / 3,
alpha=alp)
plt.annotate(' {} mil.'.format(round(W0E1R1.mean(), 1)),
xy=(W0E1R1.mean(), 3 + wid / 2),
va='center',
ha='left',
color=greys_pal[6])
plt.barh(2,
W1E0R1.mean(),
facecolor=pal['nonag_wage'],
lw=lw,
height=wid * 2 / 3,
alpha=alp)
plt.barh(2 + wid / 3,
W1E0R1.mean(),
facecolor=pal['remits'],
lw=lw,
height=wid * 2 / 3,
alpha=alp)
plt.annotate(' {} mil.'.format(round(W1E0R1.mean(), 1)),
xy=(W1E0R1.mean(), 2 + wid / 2),
va='center',
ha='left',
color=greys_pal[6])
plt.barh(1,
W1E1R0.mean(),
facecolor=pal['nonag_wage'],
lw=lw,
height=wid * 2 / 3,
alpha=alp)
plt.barh(1 + wid / 3,
W1E1R0.mean(),
facecolor=pal['entrep'],
lw=lw,
height=wid * 2 / 3,
alpha=alp)
plt.annotate(' {} mil.'.format(round(W1E1R0.mean(), 1)),
xy=(W1E1R0.mean(), 1 + wid / 2),
va='center',
ha='left',
color=greys_pal[6])
# plt.bar(9,base.mean(),facecolor=pal['remits'],lw=lw,width=wid/3,alpha=alp)
# plt.bar(9+wid/3,base.mean(),facecolor=pal['entrep'],lw=lw,width=wid/3,alpha=alp)
# plt.bar(9+wid*2/3,base.mean(),facecolor=pal['nonag_wage'],lw=lw,width=wid/3,alpha=alp)
plt.barh([1, 2, 3, 5, 6, 7],
6 * [base.mean()],
facecolor="None",
lw=1,
height=wid,
edgecolor=greys_pal[7],
alpha=0.3)
plt.grid(True, axis='x', alpha=0.3)
# plt.legend(loc='upper left',labelspacing=0.75,ncol=1,fontsize=8,borderpad=0.75,fancybox=True,frameon=True,framealpha=0.9)
plt.yticks([_ + wid / 2 for _ in range(1, 8)], [
'wages &\nentrepreneurial', 'wages &\nremittances',
'entrepreneurial &\n remittances', '', 'wage\neffect',
'entrepreneurial\neffect', 'remittance\neffect'
],
va='center',
ha='right')
plt.xlabel('Poverty increase [mil.]', labelpad=10)
plt.ylim(0.75, 8.05)
sns.despine()
plt.savefig('figs/build_a_crisis.pdf', format='pdf', bbox_inches='tight')
plt.close('all')
def plot_income_hist(hh_df, shock_code='base', use_expenditures=False):
mc = monte_carlo(0, 'base')
###############################################
# function can plot income or expenditures
# --> based on flag use_expenditures above
_fom = 'inc'
_label = 'Income'
if use_expenditures:
_fom = 'exp'
_label = 'Expenditures'
# use central value of MC series
ts = load_consumption_time_series(shock_code)
classes = {
'sub': [ts[0].T, -1E9, 1.9],
'pov': [ts[1].T, 1.9, 3.2],
'vul': [ts[2].T, 3.2, 5.5],
'sec': [ts[3].T, 5.5, 15.],
'mc': [ts[4].T, 15., 1E9]
}
###############################################
# upper limit, binning of histogram
_ul = 20
nbins = int(50)
# Income dist before disaster
ci_hgt, _bins = np.histogram(
(mc.m2d * hh_df['pc' + _fom + '_initial']).clip(upper=_ul),
bins=nbins,
weights=hh_df['popwgt'])
wid = (_bins[1] - _bins[0]) / 2
# Income dist after disaster
cf_hgt, _ = np.histogram(
(mc.m2d * hh_df['pc' + _fom + '_final']).clip(upper=_ul),
bins=_bins,
weights=hh_df['popwgt'])
# plot them
ax = plt.bar(_bins[:-1],
ci_hgt,
width=wid,
align='edge',
linewidth=0,
alpha=0.4,
facecolor=cool_pal[5],
label='2015 FIES')
ax = plt.bar(_bins[:-1] - wid,
cf_hgt,
width=wid,
align='edge',
linewidth=0,
alpha=0.4,
facecolor=cool_pal[0],
label='COVID shock')
# annotate shifts
pop_aff, frac_aff, tot_loss, di_tot, di_aff, affpop_sub, affpop_pov, affpop_vul = load_income_impacts(
shock_code)
_y = 0.0
_dy = {'sub': 9.2, 'pov': 7.8, 'vul': 6.3, 'sec': 3.7, 'mc': 0.5}
lbl = {
'sub': 'extreme poverty',
'pov': 'poverty',
'vul': 'vulnerable',
'sec': 'secure',
'mc': 'middle class'
}
for _c in ['sub', 'pov', 'vul', 'sec', 'mc']:
_mc, _min, _max = classes[_c]
# segment plot
# if _c != 'sub':
plt.plot([max(0.05, _min), max(0.05, _min)], [0, _dy[_c] + 1.0],
color=greys_pal[6],
lw=1.4,
ls='-',
clip_on=False)
plt.plot([max(0.05, _min), max(0.05, _min) + 0.1],
[_dy[_c] + 1.0, _dy[_c] + 1.10],
color=greys_pal[6],
lw=1.4,
ls='-',
clip_on=False)
# annotate population shifts
# This loads population values from MC (consumption series)
if not use_expenditures:
print(
'\n\nusing consumption (net sav), month = 12 for income hist')
tot_pop = hh_df['popwgt'].sum()
popi = round(1E-2 * tot_pop * _mc[0].mean(), 1)
popf = round(1E-2 * tot_pop * _mc[12].mean(), 1)
popf_min = round(1E-2 * tot_pop * _mc[12].quantile(0.25),
1) # popf-pop_err
popf_max = round(1E-2 * tot_pop * _mc[12].quantile(0.75),
1) # popf+pop_err
# get delta
dp = round(popf - popi, 1)
dp_max = round(popf_max - popi, 1)
dp_min = round(popf_min - popi, 1)
# pop frac affected
frac_aff_min = int(round(frac_aff[_c].quantile(0.25)))
frac_aff_max = int(round(frac_aff[_c].quantile(0.75)))
if frac_aff_min != frac_aff_max:
_anno = r'{}% $\endash$ {}%'.format(frac_aff_min, frac_aff_max)
else:
_anno = r'{}%'.format(int(round(frac_aff[_c].mean())))
_anno += ' of {} m. affected'.format(popi) + '\n'
_anno += 'final pop: {} $\endash$ {} m.\n'.format(popf_min, popf_max)
_anno += 'net shift: {}{} $\endash$ {}{} m.'.format(
'+' if dp_min > 0 else '', dp_min, '+' if dp_max > 0 else '',
dp_max)
# income loss among affected
# pcinc_init = mc.m2d*(hh_df.loc[hh_df['initial_class']==_c,['pcinc_initial','popwgt']].prod(axis=1).sum()
# /hh_df.loc[hh_df['initial_class']==_c,'popwgt'].sum())
# di_aff_min = round(mc.m2d*di_aff[_c].quantile(0.25),1)
# di_aff_max = round(mc.m2d*di_aff[_c].quantile(0.75),1)
# if di_aff_min == di_aff_max :
# _anno += r'\${}0{}'.format(round(mc.m2d*di_aff[_c].mean(),1),'/cap/day')
# _anno += r' ({}%)'.format(int(1E2*mc.m2d*di_aff[_c].mean()/pcinc_init))+'lost \n'
# else:
# _anno += r'\${}0 $\endash$ \${}0{}'.format(di_aff_min,di_aff_max,'/cap/day')
# _anno += r' ({}% $\endash$ {}%)'.format(int(1E2*di_aff_min/pcinc_init),int(1E2*di_aff_max/pcinc_init))+' lost\n'
# if use_expenditures:
plt.annotate(lbl[_c],
xy=(max(0, _min) + 0.3, _y + _dy[_c] + 1.05),
ha='left',
va='bottom',
fontsize=8,
color=greys_pal[7],
annotation_clip=False,
weight='bold')
plt.annotate(_anno,
xy=(max(0, _min) + .55, _y + _dy[_c] + 0.94),
ha='left',
va='top',
fontsize=8,
color=greys_pal[7],
annotation_clip=False,
linespacing=1.5)
plt.xlabel(_label + ' [PPP$/cap/day]', labelpad=10)
plt.ylabel('Population [millions]', labelpad=10)
plt.xlim(0)
plt.ylim(0, 8.5)
plt.yticks([n for n in range(1, 9)])
plt.grid(False)
plt.legend(labelspacing=0.75,
ncol=1,
fontsize=8,
borderpad=0.75,
fancybox=True,
frameon=True,
framealpha=0.9)
sns.despine(left=True)
plt.savefig('figs/{}_hist.pdf'.format(_label.lower()),
format='pdf',
bbox_inches='tight')
plt.close('all')
def plot_income_loss_by_channel(scode):
pal = monte_carlo(0, 'base').ichan_cols
_w = 0.75
_fs = 8
#flag
fig, ax = plt.subplots(nrows=1, ncols=2, figsize=(8, 5))
plt.axes(ax[0])
rel = 'loss' # 'lossfrac' returns fraction *of each channel*
agwage = pd.read_csv('monte_carlo/{}/ag_wage_{}.csv'.format(scode, rel),
index_col=0)
nagwage = pd.read_csv('monte_carlo/{}/nonag_wage_{}.csv'.format(
scode, rel),
index_col=0)
remits = pd.read_csv('monte_carlo/{}/remits_{}.csv'.format(scode, rel),
index_col=0)
entrep = pd.read_csv('monte_carlo/{}/entrep_{}.csv'.format(scode, rel),
index_col=0)
#
tot_loss = pd.read_csv('monte_carlo/{}/tot_loss.csv'.format(scode),
index_col=0)
# load initial income [mil. PPP/month]
init_inc_aff = pd.read_csv(
'monte_carlo/{}/tot_inc_initial_affected.csv'.format(scode),
index_col=0)
# load affected population
aff_pop = pd.read_csv('monte_carlo/{}/pop_aff.csv'.format(scode),
index_col=0)
# load total population
# plot losses in PPP/cap
btm = [0 for _ in agwage.columns]
for ncl, cl in enumerate(['sub', 'pov', 'vul', 'sec', 'mc']):
plt.bar(ncl,
nagwage[cl].mean() / aff_pop[cl].mean(),
bottom=btm[ncl],
color=pal['nonag_wage'],
width=_w,
alpha=0.6,
linewidth=0,
label=('wages (non-ag)' if ncl == 0 else ''))
btm[ncl] += nagwage[cl].mean() / aff_pop[cl].mean()
plt.bar(ncl,
entrep[cl].mean() / aff_pop[cl].mean(),
bottom=btm[ncl],
color=pal['entrep'],
width=_w,
alpha=0.6,
linewidth=0,
label=('entrepreneurial' if ncl == 0 else ''))
btm[ncl] += entrep[cl].mean() / aff_pop[cl].mean()
plt.bar(ncl,
remits[cl].mean() / aff_pop[cl].mean(),
bottom=btm[ncl],
color=pal['remits'],
width=_w,
alpha=0.6,
linewidth=0,
label=('intl. remittances' if ncl == 0 else ''))
btm[ncl] += remits[cl].mean() / aff_pop[cl].mean()
plt.bar(ncl,
agwage[cl].mean() / aff_pop[cl].mean(),
bottom=btm[ncl],
color=pal['ag_wage'],
width=_w,
alpha=0.6,
linewidth=0,
label=('agricultural wages' if ncl == 0 else ''))
btm[ncl] += agwage[cl].mean() / aff_pop[cl].mean()
# annotate with range
_low = int(round(tot_loss[cl].quantile(0.25) / aff_pop[cl].mean()))
_high = int(round(tot_loss[cl].quantile(0.75) / aff_pop[cl].mean()))
plt.annotate('\${}$\endash${}'.format(_low, _high),
xy=(_w / 2 + ncl, btm[ncl]),
fontsize=_fs,
color=greys_pal[7],
ha='center',
va='bottom')
plt.xticks([_w / 2 + _ for _ in range(0, 5)], [
'extreme\npoverty', 'poverty', 'vulnerable', 'secure', 'middle\nclass'
],
fontsize=_fs,
rotation=0)
plt.xlim(-0.1, 4 + _w + 0.1)
plt.yticks([_ * 50 for _ in range(0, 6)], fontsize=_fs)
plt.ylabel('Value [PPP$/cap/month]',
labelpad=10,
linespacing=1.75,
fontsize=_fs)
plt.legend(loc='upper left',
labelspacing=0.75,
ncol=1,
fontsize=_fs,
borderpad=0.75,
fancybox=True,
frameon=True,
framealpha=0.9)
plt.grid(False)
plt.grid(True, axis='y', alpha=0.4)
sns.despine(left=True)
# plt.savefig('figs/income_{}_by_channel.pdf'.format(rel),format='pdf',bbox_inches='tight')
# plt.close('all')
# plot fractional losses
plt.axes(ax[1])
btm = [0 for _ in agwage.columns]
for ncl, cl in enumerate(['sub', 'pov', 'vul', 'sec', 'mc']):
plt.bar(ncl,
1E2 * nagwage[cl].mean() / init_inc_aff[cl].mean(),
bottom=btm[ncl],
color=pal['nonag_wage'],
width=_w,
alpha=0.6,
linewidth=0,
label=('non-ag wages' if ncl == 0 else ''))
btm[ncl] += 1E2 * nagwage[cl].mean() / init_inc_aff[cl].mean()
plt.bar(ncl,
1E2 * entrep[cl].mean() / init_inc_aff[cl].mean(),
bottom=btm[ncl],
color=pal['entrep'],
width=_w,
alpha=0.6,
linewidth=0,
label=('entrepreneurial income' if ncl == 0 else ''))
btm[ncl] += 1E2 * entrep[cl].mean() / init_inc_aff[cl].mean()
plt.bar(ncl,
1E2 * remits[cl].mean() / init_inc_aff[cl].mean(),
bottom=btm[ncl],
color=pal['remits'],
width=_w,
alpha=0.6,
linewidth=0,
label=('remittances' if ncl == 0 else ''))
btm[ncl] += 1E2 * remits[cl].mean() / init_inc_aff[cl].mean()
plt.bar(ncl,
1E2 * agwage[cl].mean() / init_inc_aff[cl].mean(),
bottom=btm[ncl],
color=pal['ag_wage'],
width=_w,
alpha=0.6,
linewidth=0,
label=('ag wages' if ncl == 0 else ''))
btm[ncl] += 1E2 * agwage[cl].mean() / init_inc_aff[cl].mean()
# annotate with range
_low = int(
round(1E2 * tot_loss[cl].quantile(0.25) / init_inc_aff[cl].mean()))
_high = int(
round(1E2 * tot_loss[cl].quantile(0.75) / init_inc_aff[cl].mean()))
plt.annotate('{}$\endash${}%'.format(_low, _high),
xy=(_w / 2 + ncl, btm[ncl]),
fontsize=_fs,
color=greys_pal[7],
ha='center',
va='bottom')
plt.xticks([_w / 2 + _ for _ in range(0, 5)], [
'extreme\npoverty', 'poverty', 'vulnerable', 'secure', 'middle\nclass'
],
fontsize=_fs,
rotation=0)
plt.xlim(-0.1, 4 + _w + 0.1)
plt.yticks([_ * 10 for _ in range(0, 5)], fontsize=_fs)
plt.ylabel('Percentage of total income [%]',
labelpad=10,
linespacing=1.5,
fontsize=_fs)
#plt.legend(loc='upper left',labelspacing=0.75,ncol=1,fontsize=8,borderpad=0.75,fancybox=True,frameon=True,framealpha=0.9)
plt.grid(False)
plt.grid(True, axis='y', alpha=0.4)
#plt.grid(True,axis='y')
sns.despine(left=True)
plt.savefig('figs/income_loss_by_channel.pdf'.format(rel),
format='pdf',
bbox_inches='tight')
plt.close('all')
def plot_regional_poverty(scode):
mc = monte_carlo(0, 'base')
pov_init = pd.read_csv('monte_carlo/regional_poverty.csv',
index_col=0,
dtype='float',
header=None).fillna(0)
pov_init.columns = ['init']
pov_init.index = pov_init.index.astype('int')
#
pov_covid = pd.read_csv(
'monte_carlo/{}/regional_pov_covid.csv'.format(scode),
index_col=0,
dtype='float').mean(axis=0).T.to_frame(name='covid')
pov_covid.index = pov_covid.index.astype('int')
#
pov_ESP = pd.read_csv(
'monte_carlo/{}/regional_pov_ESP_eligerr50.csv'.format(scode),
index_col=0,
dtype='float').mean(axis=0).T.to_frame(name='ESP')
pov_ESP.index = pov_ESP.index.astype('int')
#
pov = pd.concat([pov_init, pov_covid, pov_ESP], axis=1)
pov.index.name = 'region'
pov = pov.reset_index()
prov_code, reg_code = get_places_dict('PH')
pov['region'].replace(reg_code, inplace=True)
pov = pov.reset_index(drop=True).set_index('region')
pov = pov.sort_values(by='ESP', ascending=True)
pov.to_csv('csv/regional_poverty.csv')
for _n, _ in enumerate(pov.index):
_lbl = 'pre-COVID (total = {} mil.)'.format(round(
pov['init'].sum(), 1)) if _n == 0 else ''
plt.plot([3 * _n, 3 * _n + 2],
[pov.loc[_, 'init'], pov.loc[_, 'init']],
lw=1,
color=greys_pal[6],
zorder=90,
label=_lbl)
plt.bar([3 * _ for _ in range(len(pov.index))],
pov['covid'],
facecolor=mc.sns_pal[2],
lw=0,
width=1,
alpha=0.6,
label='COVID shock ({} mil.)'.format(round(pov['covid'].sum(), 1)),
zorder=80)
plt.bar([3 * _ + 1 for _ in range(len(pov.index))],
pov['ESP'],
facecolor=mc.sns_pal[1],
lw=0,
width=1,
alpha=0.6,
label='with SAP benefits ({} mil.)'.format(
round(pov['ESP'].sum(), 1)),
zorder=80)
plt.grid(True, axis='y', alpha=0.3)
plt.legend(loc='upper left',
labelspacing=0.75,
ncol=1,
fontsize=8,
borderpad=0.75,
fancybox=True,
frameon=True,
framealpha=0.9)
plt.xticks([3 * _ + 1 for _ in range(len(pov.index))],
pov.index,
rotation=90)
plt.ylabel('Poverty incidence [mil.]', labelpad=10)
sns.despine(left=True)
plt.savefig('figs/regional_poverty.pdf', format='pdf', bbox_inches='tight')
plt.close('all')
def plot_income_profile(hh_df, affected_only=False):
pal = monte_carlo().ichan_cols
_ul = 500
nbins = int(40)
# total income <-- use to structure plot
tot_hgt, _bins = np.histogram(hh_df.eval('pcinc_initial').clip(upper=_ul),
bins=nbins,
weights=hh_df['popwgt'])
wid = (_bins[1] - _bins[0])
frac_nonag_hgt = []
frac_ag_hgt = []
frac_totent_hgt = []
frac_public_hgt = []
frac_remit_hgt = []
frac_dom_remit_hgt = []
for n, b in enumerate(_bins):
try:
bin_slice = '(pcinc_initial>@b)&(pcinc_initial<=@_bins[@n+1]){}'.format(
'&(income_loss>0)' if affected_only else '')
frac_nonag_hgt.append(
float(hh_df.loc[hh_df.eval(bin_slice)].eval(
'1E2*popwgt*nonagri_sal').sum() / hh_df.loc[hh_df.eval(
bin_slice)].eval('popwgt*hhinc').sum()))
frac_ag_hgt.append(
float(hh_df.loc[hh_df.eval(bin_slice)].eval(
'1E2*popwgt*agri_sal').sum() / hh_df.loc[hh_df.eval(
bin_slice)].eval('popwgt*hhinc').sum()))
frac_public_hgt.append(
float(hh_df.loc[hh_df.eval(bin_slice)].eval(
'1E2*popwgt*total_public').sum() / hh_df.loc[hh_df.eval(
bin_slice)].eval('popwgt*hhinc').sum()))
frac_remit_hgt.append(
float(hh_df.loc[hh_df.eval(bin_slice)].eval(
'1E2*popwgt*cash_abroad').sum() / hh_df.loc[hh_df.eval(
bin_slice)].eval('popwgt*hhinc').sum()))
frac_dom_remit_hgt.append(
float(hh_df.loc[hh_df.eval(bin_slice)].eval(
'1E2*popwgt*cash_domestic').sum() / hh_df.loc[hh_df.eval(
bin_slice)].eval('popwgt*hhinc').sum()))
frac_totent_hgt.append(
float(hh_df.loc[hh_df.eval(bin_slice)].eval(
'1E2*popwgt*total_entrepreneurial').sum() / hh_df.loc[
hh_df.eval(bin_slice)].eval('popwgt*hhinc').sum()))
except:
pass
_btm = [0 for _ in frac_remit_hgt]
for istream in [[frac_totent_hgt, 'entrepreneurial', pal['entrep']],
[frac_nonag_hgt, 'wages (non-ag)', pal['nonag_wage']],
[frac_ag_hgt, 'agricultural wages', pal['ag_wage']],
[frac_public_hgt, 'public transfers', pal['pub_trans']],
[frac_remit_hgt, 'intl. remittances', pal['remits']],
[
frac_dom_remit_hgt, 'domestic remittances',
pal['dom_remits']
]]:
ax = plt.bar(_bins[:-1],
istream[0],
bottom=_btm,
width=wid,
align='edge',
linewidth=0,
alpha=0.6,
facecolor=istream[2],
label=istream[1])
_btm = [i + j for i, j in zip(_btm, istream[0])]
plt.legend(loc='center right',
labelspacing=0.75,
ncol=1,
fontsize=8,
borderpad=0.75,
fancybox=True,
frameon=True,
framealpha=0.9)
plt.xlabel('pre-COVID income [PPP$/cap/month]', labelpad=8)
plt.ylabel('fraction of total income [%]', labelpad=8)
plt.xlim(0)
plt.ylim(0, 100)
plt.grid(True, axis='y', alpha=0.3)
sns.despine(left=True)
#plt.plot([0,_ul],[0,0],color=greys_pal[4],lw=1)
plt.savefig('figs/income_composition{}.pdf'.format(
'_affected_only' if affected_only else ''),
format='pdf',
bbox_inches='tight')
plt.close('all')
def plot_sectoral_income(plot_losses=False):
fig, ax = plt.subplots(nrows=1, ncols=2, figsize=(10, 7))
df = pd.read_csv('monte_carlo/sectoral_income_by_decile.csv',
index_col=['decile', 'LFS_sector']).sort_index()
df[['nonagri_sal', 'agri_sal', 'nonag_wage_loss', 'ag_wage_loss']] *= 1E-3
df_sum = df.sum(level='LFS_sector')
df_sum.loc['total'] = df_sum.sum(axis=0)
# split into 2 dfs: categories > (threshold)%, and all others
threshold = 0.07
df_sum_other = df_sum.drop(
'ag', axis=0).loc[df_sum['nonagri_sal'] < threshold *
float(df_sum.loc['total', 'nonagri_sal'])]
# get categories > (threshold)%
df_sum = df_sum.loc[df_sum['nonagri_sal'] >= threshold *
float(df_sum.loc['total', 'nonagri_sal'])].drop(
'total', axis=0)
df_sum.loc['remaining categories'] = df_sum_other.sum(axis=0)
# put this info back into decile-level df
df_other = df.reset_index().set_index('LFS_sector').loc[df_sum_other.index]
df_other = df_other.reset_index().set_index(['decile', 'LFS_sector'
]).sum(level='decile')
df_other['LFS_sector'] = 'remaining categories'
# drop other sectors from df
df = df.reset_index().set_index('LFS_sector').drop(df_sum_other.index,
axis=0)
# merge
df = pd.concat([df.reset_index(), df_other.reset_index()],
ignore_index=True).set_index(['decile', 'LFS_sector'
]).sort_index().reset_index()
# merge ag & non-ag columns
df['wage_loss'] = df['nonag_wage_loss'].copy()
df.loc[df.LFS_sector == 'ag', 'wage_loss'] = df.loc[df.LFS_sector == 'ag',
'ag_wage_loss']
df['total_wages'] = df['nonagri_sal'].copy()
df.loc[df.LFS_sector == 'ag',
'total_wages'] = df.loc[df.LFS_sector == 'ag', 'agri_sal']
df.drop(['nonag_wage_loss', 'ag_wage_loss', 'nonagri_sal', 'agri_sal'],
axis=1,
inplace=True)
# calculate sectoral contribution to wage income, by decile
df['sector_frac_total_wages'] = df['total_wages'] / df.groupby(
'decile')['total_wages'].transform('sum')
df['sector_losses_frac_total_wages'] = df['wage_loss'] / df.groupby(
'decile')['total_wages'].transform('sum')
# plotz
# left subplot: total value of each sector
plt.axes(ax[0])
output_col = 'wage_loss' if plot_losses else 'total_wages'
bot = [0 for _ in range(1, 11)]
sectors = df.LFS_sector.unique()
for _n, _sec in enumerate(
np.append(sectors[sectors != 'remaining categories'],
'remaining categories')):
try:
sec_lbl = monte_carlo().sector_labels[_sec]
except:
sec_lbl = _sec
ddf = df.loc[df['LFS_sector'] == _sec]
plt.bar(ddf['decile'],
ddf[output_col],
bottom=bot,
label=sec_lbl,
facecolor=sns_pal[_n],
alpha=0.7,
lw=0)
bot += np.array(ddf[output_col].squeeze().T)
plt.xlim(0.75, 11)
plt.xticks([_ + 0.4 for _ in range(1, 11)],
range(1, 11),
linespacing=0.80,
fontsize=9,
ha='center')
plt.xlabel('Income decile', fontsize=9, labelpad=10)
plt.yticks([0.5 * _ for _ in range(1, 8)],
linespacing=0.80,
fontsize=9,
ha='center')
plt.ylabel('Wage income {}[billion PPP\$]'.format(
'losses ' if plot_losses else ''),
fontsize=9,
labelpad=10)
plt.grid(True, axis='y', alpha=0.3)
plt.legend(loc='upper left',
labelspacing=0.75,
ncol=1,
fontsize=8,
borderpad=0.75,
fancybox=True,
frameon=True,
framealpha=0.9)
sns.despine(left=True)
# right subplot: relative contribution of each sector to total wage inome
plt.axes(ax[1])
output_col = 'sector_losses_frac_total_wages' if plot_losses else 'sector_frac_total_wages'
bot = [0 for _ in range(1, 11)]
sectors = df.LFS_sector.unique()
for _n, _sec in enumerate(
np.append(sectors[sectors != 'remaining categories'],
'remaining categories')):
try:
sec_lbl = monte_carlo().sector_labels[_sec]
except:
sec_lbl = _sec
ddf = df.loc[df['LFS_sector'] == _sec]
plt.bar(ddf['decile'],
ddf[output_col],
bottom=bot,
label=sec_lbl,
facecolor=sns_pal[_n],
alpha=0.7,
lw=0)
bot += np.array(ddf[output_col].squeeze().T)
plt.xlim(0.75, 11)
plt.xticks([_ + 0.4 for _ in range(1, 11)],
range(1, 11),
linespacing=0.80,
fontsize=9,
ha='center')
plt.xlabel('Income decile', fontsize=9, labelpad=10)
plt.ylim(0, 1)
plt.yticks([0.2 * _ for _ in range(1, 6)],
linespacing=0.80,
fontsize=9,
ha='center')
plt.ylabel('Fraction of total wage income {}[%]'.format(
'lost ' if plot_losses else ''),
fontsize=9,
labelpad=10)
plt.grid(True, axis='y', alpha=0.3)
plt.savefig('figs/wage_income_sectoral_{}.pdf'.format(
'losses' if plot_losses else 'composition'),
format='pdf',
bbox_inches='tight')
plt.close('all')
def plot_losses_total_value(scode):
plt.gcf().set_size_inches(6, 8)
barwid = 2
##########################################
# LOAD RESULTS
# wage sector total value
wage_val = pd.read_csv(
'monte_carlo/{}/total_value_wages.csv'.format(scode),
index_col=0).multiply(1E-3)
tot_wage_val = wage_val.sum(axis=1).mean()
# wage sector losses
wage_loss = pd.read_csv(
'monte_carlo/{}/total_loss_wages.csv'.format(scode),
index_col=0).multiply(1E-3)
tot_wage_loss = wage_loss.sum(axis=1).mean()
#####
# entrepreneurial total value
ent_val = pd.read_csv('monte_carlo/{}/total_value_ent.csv'.format(scode),
index_col=0).multiply(1E-3)
tot_ent_val = ent_val.sum(axis=1).mean()
# entrepreneurial losses
ent_loss = pd.read_csv('monte_carlo/{}/total_loss_ent.csv'.format(scode),
index_col=0).multiply(1E-3)
tot_ent_loss = ent_loss.sum(axis=1).mean()
#####
# remittances total value
remits_value = pd.read_csv(
'monte_carlo/{}/total_value_remits.csv'.format(scode),
index_col=0).multiply(1E-3)
intl_remits_val = remits_value['intl'].mean()
# remittances losses
remits_loss = pd.read_csv(
'monte_carlo/{}/total_loss_remits.csv'.format(scode),
index_col=0).multiply(1E-3)
tot_remits_loss = remits_loss.sum(axis=1).mean()
#####
# total economic value
economic_value = pd.read_csv(
'monte_carlo/{}/total_value_economy.csv'.format(scode),
index_col=0).multiply(1E-3)
total_economic_val = economic_value['income'].mean()
total_economic_loss = economic_value['loss'].mean()
# concat & transpose
#value = pd.concat([wage_val,ent_val], axis=1)
loss = pd.concat([wage_loss, ent_loss, remits_loss], axis=1).T
loss['avg'] = loss.mean(axis=1)
loss = loss.loc[loss.avg > 5E-2 * loss['avg'].max()]
loss = loss.sort_values(by='avg', ascending=True).drop('avg', axis=1).T
# coloring
# pal = monte_carlo(0,'base').ichan_cols
# colors = [pal['nonag_wage'] if _ in wage_loss.columns else (pal['entrep'] if _ in ent_loss.columns else pal['remits']) for _ in loss.columns]
# joyplot
min_loss = loss.min(axis=0)
d_loss = loss.max(axis=0) - loss.min(axis=0)
mean_loss = loss.mean(axis=0)
joyplot_loss = loss.T.stack().reset_index()
joyplot_loss.columns = ['wage_sector', 'nsim', 'result']
joyplot_grouped = joyplot_loss.groupby('wage_sector', sort=False)
sector_labels = monte_carlo(0, 'base').sector_labels
_ylabelsA = [
_ + ' (w)' if _ in wage_loss.columns else
(_ + ' (e)' if _ in ent_loss.columns else _ + 'xxxx')
for _ in loss.columns
]
_ylabels = [(sector_labels[_[:-4]] + _[-4:]).lower().replace('xxxx', '')
if _[:-4] in sector_labels else _.lower() for _ in _ylabelsA]
fig, axes = joypy.joyplot(joyplot_grouped,
column='result',
by='wage_sector',
overlap=3,
x_range=[0, 1.20],
grid='y',
figsize=(10, 10),
linewidth=0.02,
alpha=0.8,
ylabelsize=15,
xlabelsize=15,
labels=_ylabels,
range_style='own',
colormap=cm.RdYlGn_r)
for n, y in enumerate(loss.columns):
if mean_loss[y] != 0:
axes[n].plot(
[loss.quantile(.25, axis=0)[y],
loss.quantile(.25, axis=0)[y]], [-2, 2],
lw=1.,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].plot(
[loss.quantile(.75, axis=0)[y],
loss.quantile(.75, axis=0)[y]], [-2, 2],
lw=1.0,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].plot(
[loss.quantile(.25, axis=0)[y],
loss.quantile(.75, axis=0)[y]], [0, 0],
lw=1.0,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].annotate(
round(mean_loss[y], 2),
xy=(mean_loss[y],
0.05 - (-1 if _ylabels[n] == 'wholesale (w)' else 1.5)),
fontsize=9,
ha='center',
annotation_clip=False,
zorder=100,
va=('bottom' if _ylabels[n] == 'wholesale (w)' else 'top'),
color=("white"
if _ylabels[n] == 'wholesale (w)' else greys_pal[7]))
else:
plt.annotate('no impact',
xy=(mean_loss[y] + 0.95, n + 0.9),
fontsize=7,
ha='left',
va='center',
weight='bold',
zorder=92,
style='italic')
# loss = loss.T.stack().reset_index()
# loss.columns = ['sector','sim','loss_value']
# Editing /Users/brian/Software/anaconda3/lib/python3.5/site-packages/seaborn/categorical.py (draw_box_lines) to customize
# sns.violinplot(x='loss_value', y='sector', data=loss, scale='count', inner='box',cut=0,palette=colors,saturation=0.6,linewidth=0.05)
# sectoral labels
sector_labels = monte_carlo(0, 'base').sector_labels
seclabels = [
sector_labels[_] if _ in sector_labels else _.lower()
for _ in loss.columns
]
plt.yticks([barwid / 2 + _ * 1.4 for _ in range(len(loss.columns))],
seclabels,
linespacing=0.80,
fontsize=9)
# totals
plt.annotate('Monthly losses',
xy=(0.655, 0.868),
xycoords='axes fraction',
ha='left',
va='top',
fontsize=14,
weight='bold')
anno1 = round(1E2 * economic_value['loss'].quantile(.25) / 699.258, 1)
anno2 = round(1E2 * economic_value['loss'].quantile(.75) / 699.258, 1)
annostr = (r'PPP\${} $\endash$ {} bil.'.format(
round(economic_value['loss'].quantile(.25), 1),
round(economic_value['loss'].quantile(.75), 1)) + '\n' +
r'{}%$\endash${}% of total income'.format(
round(
1E2 * (economic_value['loss'].quantile(.25)) /
total_economic_val, 1),
round(
1E2 * (economic_value['loss'].quantile(.75)) /
total_economic_val, 1), round(total_economic_val, 1)) +
'\n' + r'{}%$\endash${}% of GDP per month'.format(anno1, anno2))
plt.annotate(annostr,
xy=(0.985, 0.838),
xycoords='axes fraction',
ha='right',
va='top',
fontsize=12,
linespacing=1.25,
zorder=100)
# wages
annostr = ('(w) wage sectors:\nPPP\${} $\endash$ {} bil.'.format(
round(wage_loss.sum(axis=1).quantile(0.25), 1),
round(wage_loss.sum(axis=1).quantile(0.75), 1)
) + r' ({}%$\endash${}%)'.format(
int(round(1E2 * wage_loss.sum(axis=1).quantile(0.25) / tot_wage_val)),
int(round(1E2 * wage_loss.sum(axis=1).quantile(0.75) / tot_wage_val)),
round(tot_wage_val, 1)))
plt.annotate(annostr,
xy=(0.655, 0.758),
xycoords='axes fraction',
ha='left',
va='top',
fontsize=12,
linespacing=1.25,
zorder=100)
# entrepreneurial
annostr = (
'(e) entrepreneurial sectors:\nPPP\${} $\endash$ {} bil.'.format(
round(ent_loss.sum(axis=1).quantile(0.25), 1),
round(ent_loss.sum(axis=1).quantile(0.75), 1)) +
r' ({}%$\endash${}%)'.format(
int(round(
1E2 * ent_loss.sum(axis=1).quantile(0.25) / tot_ent_val)),
int(round(1E2 * ent_loss.sum(axis=1).quantile(0.75) /
tot_ent_val)), round(tot_ent_val, 1)))
plt.annotate(annostr,
xy=(0.655, 0.700),
xycoords='axes fraction',
ha='left',
va='top',
fontsize=12,
linespacing=1.25,
zorder=100)
# remittances
q25a = round(remits_loss['intl'].quantile(0.25), 1)
q75a = round(remits_loss['intl'].quantile(0.75), 1)
q25b = int(
round(1E2 * remits_loss['intl'].quantile(0.25) / intl_remits_val))
q75b = int(
round(1E2 * remits_loss['intl'].quantile(0.75) / intl_remits_val))
if q25a != q75a:
annostr = (
'international remittances:\nPPP\${} $\endash$ {} bil.'.format(
q25a, q75a) + r' ({}%$\endash${}%)'.format(q25b, q75b))
else:
annostr = ('international remittances:\nPPP\${} bil.'.format(q25a) +
r' ({}%)'.format(q25b))
plt.annotate(annostr,
xy=(0.655, 0.643),
xycoords='axes fraction',
ha='left',
va='top',
fontsize=12,
linespacing=1.25,
zorder=100)
# need to add bbox behind annotations
# x-ax label
plt.xlabel('Aggregate loss [billion PPP$/month]', labelpad=10, fontsize=15)
plt.grid(True, axis='y', alpha=0.5, zorder=80)
sns.despine(left=True)
plt.savefig('figs/sectoral_losses_total_value.pdf',
format='pdf',
bbox_inches='tight')
plt.close('all')
def plot_sectoral_impacts(scode):
pal = monte_carlo().ichan_cols
###################################################
# wage sectors (LFS)
mc_floss = pd.read_csv('monte_carlo/{}/frac_loss_wages.csv'.format(scode),
index_col=0).T
try:
mc_floss = mc_floss.drop('unemployed', axis=0)
except:
pass
try:
mc_floss = mc_floss.drop('unclassified', axis=0)
except:
pass
mc_floss['avg'] = mc_floss.mean(axis=1)
mc_floss.sort_values(by='avg', ascending=True, inplace=True)
mc_floss = mc_floss.drop(['avg'], axis=1).T
#mc_loss = pd.read_csv('monte_carlo/{}/total_loss_wages.csv'.format(scode),index_col=0)
#mc_value = pd.read_csv('monte_carlo/{}/total_value_wages.csv'.format(scode),index_col=0)
min_floss = mc_floss.min(axis=0)
d_floss = mc_floss.max(axis=0) - mc_floss.min(axis=0)
mean_floss = mc_floss.mean(axis=0)
joyplot_floss = mc_floss.T.stack().reset_index()
joyplot_floss.columns = ['wage_sector', 'nsim', 'result']
joyplot_grouped = joyplot_floss.groupby('wage_sector', sort=False)
sector_labels = monte_carlo(0, 'base').sector_labels
_ylabels = [
sector_labels[_] if _ in sector_labels else _.lower()
for _ in mc_floss.columns
]
fig, axes = joypy.joyplot(joyplot_grouped,
column='result',
by='wage_sector',
overlap=3,
x_range=[0, 100],
grid='y',
figsize=(10, 10),
colormap=cm.RdYlGn_r,
linewidth=0.02,
alpha=0.8,
ylabelsize=15,
xlabelsize=15,
labels=_ylabels,
range_style='own')
up_cats = ['government', 'information', 'agriculture']
for n, y in enumerate(mc_floss.columns):
if mean_floss[y] != 0:
axes[n].plot([
mc_floss.quantile(.25, axis=0)[y],
mc_floss.quantile(.25, axis=0)[y]
], [-0.01, 0.01],
lw=1.,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].plot([
mc_floss.quantile(.75, axis=0)[y],
mc_floss.quantile(.75, axis=0)[y]
], [-0.01, 0.01],
lw=1.0,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].plot([
mc_floss.quantile(.25, axis=0)[y],
mc_floss.quantile(.75, axis=0)[y]
], [0, 0],
lw=1.0,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].annotate(
'{}'.format(int(round(mean_floss[y]))),
xy=(mean_floss[y],
0.005 - (-0.00 if _ylabels[n] in up_cats else 0.01)),
fontsize=9,
ha='center',
annotation_clip=False,
zorder=100,
va=('bottom' if _ylabels[n] in up_cats else 'top'),
color=("white" if _ylabels[n] in up_cats else greys_pal[7]))
# axes[n].plot([mean_floss[y],mean_floss[y]],[0,0.055],lw=2.0,color=greys_pal[7],zorder=100)
# axes[n].annotate('{}%'.format(int(round(mean_floss[y]))),xy=(mean_floss[y]+1,0.025),fontsize=10,ha='left',va='top',weight='bold',zorder=92,color=greys_pal[8])
else:
plt.annotate('no impact',
xy=(mean_floss[y] + 0.95, n + 0.9),
fontsize=7,
ha='left',
va='center',
weight='bold',
zorder=92,
style='italic')
# plt.yticks([0.9+_ for _ in range(0,len(mc_floss.columns))],_ylabels)
plt.xticks([0, 20, 40, 60, 80, 100],
['0%', '20%', '40%', '60%', '80%', '100%'])
plt.xlabel('Total income loss, by wage sector', labelpad=10, fontsize=15)
plt.grid(False) #True,axis='x',alpha=1.0)
sns.despine(left=True, bottom=True)
plt.savefig('figs/sectoral_losses_wage.pdf',
format='pdf',
bbox_inches='tight')
plt.close('all')
###################################################
# entrepreneurial sectors (FIES-prescribed sectors)
###################################################
# wage sectors (LFS)
mc_floss = pd.read_csv('monte_carlo/{}/frac_loss_ent.csv'.format(scode),
index_col=0).T
mc_floss['avg'] = mc_floss.mean(axis=1)
mc_floss.sort_values(by='avg', ascending=True, inplace=True)
mc_floss = mc_floss.drop(['avg'], axis=1).T
#mc_loss = pd.read_csv('monte_carlo/{}/total_loss_wages.csv'.format(scode),index_col=0)
#mc_value = pd.read_csv('monte_carlo/{}/total_value_wages.csv'.format(scode),index_col=0)
min_floss = mc_floss.min(axis=0)
d_floss = mc_floss.max(axis=0) - mc_floss.min(axis=0)
mean_floss = mc_floss.mean(axis=0)
joyplot_floss = mc_floss.T.stack().reset_index()
joyplot_floss.columns = ['wage_sector', 'nsim', 'result']
joyplot_grouped = joyplot_floss.groupby('wage_sector', sort=False)
sector_labels = monte_carlo(0, 'base').sector_labels
_ylabels = [
sector_labels[_] if _ in sector_labels else _.lower()
for _ in mc_floss.columns
]
fig, axes = joypy.joyplot(joyplot_grouped,
column='result',
by='wage_sector',
overlap=3,
x_range=[0, 100],
grid='y',
figsize=(10, 10),
colormap=cm.RdYlGn_r,
linewidth=0.02,
alpha=0.8,
ylabelsize=15,
xlabelsize=15,
labels=_ylabels,
range_style='own')
up_cats = ['fishing', 'livestock & poultry', 'crop farming & gardening']
for n, y in enumerate(mc_floss.columns):
if mean_floss[y] != 0:
axes[n].plot([
mc_floss.quantile(.25, axis=0)[y],
mc_floss.quantile(.25, axis=0)[y]
], [-0.01, 0.01],
lw=1.,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].plot([
mc_floss.quantile(.75, axis=0)[y],
mc_floss.quantile(.75, axis=0)[y]
], [-0.01, 0.01],
lw=1.0,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].plot([
mc_floss.quantile(.25, axis=0)[y],
mc_floss.quantile(.75, axis=0)[y]
], [0, 0],
lw=1.0,
color=greys_pal[7],
zorder=99,
clip_on=False)
axes[n].annotate(
'{}'.format(int(round(mean_floss[y]))),
xy=(mean_floss[y],
0.005 - (-0.0 if _ylabels[n] in up_cats else 0.01)),
fontsize=9,
ha='center',
annotation_clip=False,
zorder=100,
va=('bottom' if _ylabels[n] in up_cats else 'top'),
color=("white" if _ylabels[n] in up_cats else greys_pal[7]))
# axes[n].plot([mean_floss[y],mean_floss[y]],[0,0.055],lw=2.0,color=greys_pal[7],zorder=100)
# axes[n].annotate('{}%'.format(int(round(mean_floss[y]))),xy=(mean_floss[y]+1,0.025),fontsize=10,ha='left',va='top',weight='bold',zorder=92,color=greys_pal[8])
else:
plt.annotate('no impact',
xy=(mean_floss[y] + 0.95, n + 0.9),
fontsize=7,
ha='left',
va='center',
weight='bold',
zorder=92,
style='italic')
# plt.yticks([0.9+_ for _ in range(0,len(mc_floss.columns))],_ylabels)
plt.xticks([0, 20, 40, 60, 80, 100],
['0%', '20%', '40%', '60%', '80%', '100%'])
plt.xlabel('Total income loss, by entrepreneurial sector',
labelpad=10,
fontsize=15)
plt.grid(False) #True,axis='x',alpha=1.0)
sns.despine(left=True, bottom=True)
plt.savefig('figs/sectoral_losses_entrepreneurial.pdf',
format='pdf',
bbox_inches='tight')
plt.close('all')