Lecture17_choropleths_finished

Lecture 17: Plotting Data with Maps

We have figured out how to plot geographic data using geopandas. We needed shape files loaded into geoDataFrames with one column assigned as the geoDataFrame's geometry. Once that is set up, plotting is easy.

In this notebook we are going to learn how to assign colors to maps based on a variable — a chloropeth. An example of this kind of map are the 2016 voting patterns in Georgia. By the end of this workbook, we will have made a similar map but with 2020 data.

Here's an outline of today's lecture:

What is a Choropleth Map?

Georgia Presidential Returns

Class Announcements

PS4 is on eLC and is due by end-of-day Sunday, October 23rd. Usual office hours via Zoom tomorrow. See syllabus for the zoom link.

1. What is a Choropeth Map? (return home)¶

A choropeth map is one with regions assigned colors based on a variable.

'column' is the column name that holds the variable to color by
'cmap' is the color scheme. There are many

Let's return to the South American map from the Maps notebook.

In [1]:

import pandas as pd                         # pandas for data management
import geopandas                            # geopandas for maps work
import matplotlib.pyplot as plt             # matplotlib for plotting details

In [2]:

# geopandas comes with some datasets that define maps. 
# Here, we grab a low-resolution Natural Earth map and load it into a GeoDataFrame.
world = geopandas.read_file(geopandas.datasets.get_path('naturalearth_lowres'))
world.sample(5)

Out[2]:

	pop_est	continent	name	iso_a3	gdp_md_est	geometry
31	32510453.0	South America	Peru	PER	226848	POLYGON ((-69.89364 -4.29819, -70.79477 -4.251...
21	5347896.0	Europe	Norway	NOR	403336	MULTIPOLYGON (((15.14282 79.67431, 15.52255 80...
125	2854191.0	Europe	Albania	ALB	15279	POLYGON ((21.02004 40.84273, 20.99999 40.58000...
67	5380508.0	Africa	Congo	COG	12267	POLYGON ((18.45307 3.50439, 18.39379 2.90044, ...
164	6777452.0	Africa	Libya	LBY	52091	POLYGON ((25.00000 22.00000, 25.00000 20.00304...

The dataset includes measures of population and GDP. Let's create a map where the color of each country depends on GDP per capita.

In [3]:

world['gdp_cap'] = world['gdp_md_est']/world['pop_est']
world['gdp_cap'].describe()

Out[3]:

count    177.000000
mean       0.016193
std        0.025676
min        0.000261
25%        0.001817
50%        0.005790
75%        0.017828
max        0.200000
Name: gdp_cap, dtype: float64

These numbers don't make much sense. There must be a scaling issue. This is why we always do sanity checks; ie, we ask "Do the data look like we expect?"

The GDP data are in millions (eg, US GDP is roughly 20 trillion). The population data are not adjusted. (US population is around 350 million).

In [4]:

print('{0:,}'.format(float(world[world['iso_a3']=='USA']['gdp_md_est'])))

21,433,226.0

In [5]:

print('{0:,}'.format(float(world[world['iso_a3']=='USA']['pop_est'])))

328,239,523.0

In [6]:

world['gdp_cap'] = world['gdp_md_est']/world['pop_est']*1000000
print('US GDP per capita is {0:,.0f}'.format(float(world[world['iso_a3']=='USA']['gdp_cap'])))

US GDP per capita is 65,298

That looks better. Let's check it out once more.

In [7]:

world.sort_values('gdp_cap', ascending=False)

Out[7]:

	pop_est	continent	name	iso_a3	gdp_md_est	geometry	gdp_cap
159	4490.0	Antarctica	Antarctica	ATA	898	MULTIPOLYGON (((-48.66062 -78.04702, -48.15140...	200000.000000
128	619896.0	Europe	Luxembourg	LUX	71104	POLYGON ((6.04307 50.12805, 6.24275 49.90223, ...	114703.111490
23	140.0	Seven seas (open ocean)	Fr. S. Antarctic Lands	ATF	16	POLYGON ((68.93500 -48.62500, 69.58000 -48.940...	114285.714286
20	3398.0	South America	Falkland Is.	FLK	282	POLYGON ((-61.20000 -51.85000, -60.00000 -51.2...	82989.994114
127	8574832.0	Europe	Switzerland	CHE	703082	POLYGON ((9.59423 47.52506, 9.63293 47.34760, ...	81993.676378
...	...	...	...	...	...	...	...
66	4745185.0	Africa	Central African Rep.	CAF	2220	POLYGON ((27.37423 5.23394, 27.04407 5.12785, ...	467.842666
12	10192317.3	Africa	Somalia	SOM	4719	POLYGON ((41.58513 -1.68325, 40.99300 -0.85829...	462.995790
71	18628747.0	Africa	Malawi	MWI	7666	POLYGON ((32.75938 -9.23060, 33.73972 -9.41715...	411.514526
154	6081196.0	Africa	Eritrea	ERI	2065	POLYGON ((36.42951 14.42211, 36.32322 14.82249...	339.571361
75	11530580.0	Africa	Burundi	BDI	3012	POLYGON ((30.46967 -2.41385, 30.52766 -2.80762...	261.218430

177 rows × 7 columns

The data for small places is crazy. For example, Antarctica's GDP per capita doesn't really make sense. Let's only plot countries with populations greater than 50,000. I'm sorry to do this to you, Falkland Islands. BTW Who owns the Falkland Islands anyways (story)?

In [8]:

world[world['pop_est']<50000]

Out[8]:

	pop_est	continent	name	iso_a3	gdp_md_est	geometry	gdp_cap
20	3398.0	South America	Falkland Is.	FLK	282	POLYGON ((-61.20000 -51.85000, -60.00000 -51.2...	82989.994114
23	140.0	Seven seas (open ocean)	Fr. S. Antarctic Lands	ATF	16	POLYGON ((68.93500 -48.62500, 69.58000 -48.940...	114285.714286
159	4490.0	Antarctica	Antarctica	ATA	898	MULTIPOLYGON (((-48.66062 -78.04702, -48.15140...	200000.000000

In [9]:

fig, gax = plt.subplots(figsize=(20,10))

world[world['pop_est']>50000].plot(ax = gax, column='gdp_cap', edgecolor='black', cmap = 'Reds', legend=False)

plt.axis('off')

plt.show()

That legend is atrocious. Never trust the defaults. Unfortunately, we are going to have to dig into the guts of matplotlib to fix it. It's a good thing we are already experienced with matplotlib.

We are going to add a separate axis to our plot. Then we will tell the geopandas plot command to add the legend to the new axis.

In [10]:

from mpl_toolkits.axes_grid1 import make_axes_locatable

fig, gax = plt.subplots(figsize=(20,10))

# Adapted from https://geopandas.org/mapping.html
# https://matplotlib.org/3.1.1/gallery/axes_grid1/demo_colorbar_with_axes_divider.html

# Create an axis divider object
divider = make_axes_locatable(gax)

# Add a second axis to the right of the original axes. 
# Make the new axes 5% of the original and add some padding between them.
legend_ax = divider.append_axes('right', size='5%', pad=0.1)


world[world['pop_est']>50000].plot(
                                   ax = gax, column='gdp_cap',           # ax is the main axes
                                   edgecolor='black', cmap = 'Reds', 
                                   legend=True, cax=legend_ax)           # cax is the legend axes
                                   
                                                
# Here I ony turn off the axis of the main plot
gax.axis('off')

plt.show()

You can experiment with different padding, widths, locations, and color maps.

2. Georgia Presidential Returns (return home)¶

Let's make the map we ended with last lecture much cooler by introducing colors to reflect voting patterns in the 2020 Presidential election.

The steps:

A. Merge data on votes with geographical data.

B. ~~Plot the borders ~~. We already did this last lecture so this part will be cake.

C. Color the map using vote shares.

A. Get the vote totals and merge them with the DataFrame containing geographical information¶

Time to add the voter totals. I downloaded the results from the Georgia Office of the Secretary of State (link). I would have called him on the phone to request a favor but...

I saved the download as GAvoting_2000.xls. The spreadsheet actually has all of the November 3, 2020 election results but we're just interested in the Presidential election which is on sheet '1'.

There's some work we need to do before we can create a choropleth with these data.

Step 1. Load 'GAvoting_2000.xls' to a DataFrame named df. Note, this is not a GeoDataFrame, this data doesn't have a geometery to it, it just has county names and vote counts.

In [11]:

df = pd.read_excel('./Data/GAvoting_2000.xls',sheet_name='1',skiprows=[0,1])
df.head()

Out[11]:

	County	Trump: Election Day Votes	Trump: Absentee by Mail Votes	Trump: Advanced Voting Votes	Trump: Provisional Votes	Trump: Total Votes	Biden: Election Day Votes	Biden: Absentee by Mail Votes	Biden: Advanced Voting Votes	Biden: Provisional Votes	Biden: Total Votes	Jorgenseon: Election Day Votes	Jorgenseon: Absentee by Mail Votes	Jorgenseon: Advanced Voting Votes	Jorgenseon: Total Votes	Total
0	APPLING	1753	890	3874	9	6526	334	587	855	3	1779	5	5	26	36	8341
1	ATKINSON	716	164	1419	1	2300	250	130	445	0	825	14	3	13	30	3155
2	BACON	431	487	3099	1	4018	140	196	288	1	625	8	4	13	25	4668
3	BAKER	291	138	466	2	897	149	234	269	0	652	2	2	2	6	1555
4	BALDWIN	1873	1290	5736	4	8903	1527	3000	4612	1	9140	63	38	107	208	18251

Step 2. Create a variable called 'trump_share' that is the share of trump votes out of the total vote count.

In [12]:

df['Trump: Share'] = 100*df['Trump: Total Votes']/df['Total']
df['Biden: Share'] = 100*df['Biden: Total Votes']/df['Total']
df['Jorgenseon: Share'] = 100*df['Jorgenseon: Total Votes']/df['Total']

Step 3. Use melt' to convert the data from wide to long. Call the new df results. Use split and partition to create a field called 'candidate' and 'category'.

In [13]:

results = pd.melt(df, id_vars=['County'], value_vars=['Trump: Election Day Votes',
                                            'Trump: Absentee by Mail Votes',
                                            'Trump: Advanced Voting Votes',
                                            'Trump: Provisional Votes',
                                            'Trump: Total Votes',
                                            'Trump: Share',
                                            'Biden: Election Day Votes',
                                            'Biden: Absentee by Mail Votes',
                                            'Biden: Advanced Voting Votes',
                                            'Biden: Provisional Votes',
                                            'Biden: Total Votes',
                                            'Biden: Share',
                                            'Jorgenseon: Election Day Votes',
                                            'Jorgenseon: Absentee by Mail Votes',
                                            'Jorgenseon: Advanced Voting Votes',
                                            'Jorgenseon: Provisional Votes',
                                            'Jorgenseon: Total Votes',
                                            'Jorgenseon: Share'
                                           ])

results['candidate'] = results['variable'].str.partition(':')[0].str.strip()
results['category'] = results['variable'].str.partition(':')[2].str.strip()
results.drop(columns='variable', inplace=True)
results.rename(columns={'value':'count'},inplace=True)
results.head()

Out[13]:

	County	count	candidate	category
0	APPLING	1753.0	Trump	Election Day Votes
1	ATKINSON	716.0	Trump	Election Day Votes
2	BACON	431.0	Trump	Election Day Votes
3	BAKER	291.0	Trump	Election Day Votes
4	BALDWIN	1873.0	Trump	Election Day Votes

The county names in the map data are in title case. The county names in the vote data are in all caps. We know how to fix this up.

Step 4. Convert the county names in results to title case.

Step 5 Strip the whitespace out of the county names in results, too. (Trust me, there is some extra space in some of the county names...) Try str.strip.

In [14]:

results['County'] = results['County'].str.title().str.strip()
results.head(5)

Out[14]:

	County	count	candidate	category
0	Appling	1753.0	Trump	Election Day Votes
1	Atkinson	716.0	Trump	Election Day Votes
2	Bacon	431.0	Trump	Election Day Votes
3	Baker	291.0	Trump	Election Day Votes
4	Baldwin	1873.0	Trump	Election Day Votes

Step 6. Load map data. Strip any whitespace from the NAME variable in the ga_counties GeoDataFrame.

Step 7. Convert the NAME variable to title case in the ga_counties GeoDataFrame.

In [15]:

counties = geopandas.read_file('./Data/cb_2017_us_county_5m/cb_2017_us_county_5m.shp')
counties['NAME'] = counties['NAME'].str.title().str.strip()

# GA fips code is 13. 
ga_counties = counties[counties['STATEFP']=='13']

# Change the projection to Mercador
ga_counties = ga_counties.to_crs({'init': 'epsg:3395'}) 

ga_counties.head()

C:\Users\jt83241\Anaconda3\envs\geo_env\lib\site-packages\pyproj\crs\crs.py:141: FutureWarning: '+init=<authority>:<code>' syntax is deprecated. '<authority>:<code>' is the preferred initialization method. When making the change, be mindful of axis order changes: https://pyproj4.github.io/pyproj/stable/gotchas.html#axis-order-changes-in-proj-6
  in_crs_string = _prepare_from_proj_string(in_crs_string)

Out[15]:

	STATEFP	COUNTYFP	COUNTYNS	AFFGEOID	GEOID	NAME	LSAD	ALAND	AWATER	geometry
8	13	143	00350637	0500000US13143	13143	Haralson	06	730804590	2616530	POLYGON ((-9505182.817 3991780.817, -9485844.5...
9	13	023	00347451	0500000US13023	13023	Bleckley	06	559100209	8447343	POLYGON ((-9293411.509 3793795.634, -9293245.8...
31	13	261	00343504	0500000US13261	13261	Sumter	06	1250167780	25799665	POLYGON ((-9400241.039 3736618.850, -9399040.1...
34	13	199	00346892	0500000US13199	13199	Meriwether	06	1298162528	10832504	POLYGON ((-9447029.178 3871454.062, -9446834.5...
90	13	097	01686467	0500000US13097	13097	Douglas	06	518261551	2517004	POLYGON ((-9451284.254 3968959.296, -9451212.6...

Step 8. Merge ga_counties and results.

In [16]:

ga_votes = pd.merge(left=ga_counties, right=results, left_on='NAME', right_on='County', how='right')
ga_votes.sample(5)

Out[16]:

	STATEFP	COUNTYFP	COUNTYNS	AFFGEOID	GEOID	NAME	LSAD	ALAND	AWATER	geometry	County	count	candidate	category
134	13	273	00352238	0500000US13273	13273	Terrell	06	8.687853e+08	5861795.0	POLYGON ((-9417603.986 3730246.439, -9401299.4...	Terrell	665.000000	Trump	Election Day Votes
924	13	253	00351263	0500000US13253	13253	Seminole	06	6.092453e+08	55231610.0	POLYGON ((-9464781.206 3620506.319, -9453520.8...	Seminole	67.224511	Trump	Share
330	13	021	01672039	0500000US13021	13021	Bibb	06	6.456998e+08	14503560.0	POLYGON ((-9338806.039 3852017.169, -9336672.0...	Bibb	13234.000000	Trump	Advanced Voting Votes
691	13	105	00347828	0500000US13105	13105	Elbert	06	9.093037e+08	59911773.0	POLYGON ((-9250147.745 4032005.602, -9248201.1...	Elbert	6226.000000	Trump	Total Votes
2628	13	139	01686953	0500000US13139	13139	Hall	06	1.017805e+09	94061469.0	POLYGON ((-9357252.792 4029432.015, -9354255.6...	Hall	1322.000000	Jorgenseon	Total Votes

B. Make the Borders¶

The code below is just recycled from the previous lecture.

In [17]:

fig, gax = plt.subplots(figsize=(10,10))

# Plot the counties
ga_counties.plot(ax=gax, edgecolor='black', color = 'white')

plt.axis('off')

plt.show()

C. Color the map¶

Create a choropeth map with colors that correspond to Trump's share of the vote.

Remember the extra arguments.

column is set to the column name of the data we want to be 'colored'
cmap determines the color scheme. I am using red colors.
legend turn on the legend

In [18]:

fig, gax = plt.subplots(1, 2, figsize = (20,10))

# TRUMP MAP
ga_votes[(ga_votes['candidate']=='Trump')&(ga_votes['category']=='Share')].plot(ax=gax[0], edgecolor='black', column='count',  legend=True, cmap='Reds', scheme='quantiles')

gax[0].set_title('Trump Vote Share',fontsize=18)

gax[0].get_legend().set_title('')
gax[0].get_legend().draw_frame(False)

# I don't want the axis with long and lat
gax[0].axis('off')

# BIDEN MAP
ga_votes[(ga_votes['candidate']=='Biden')&(ga_votes['category']=='Share')].plot(ax=gax[1], edgecolor='black', column='count',  legend=True, cmap='Blues', scheme='quantiles')

gax[1].set_title('Biden Vote Share',fontsize=18)

gax[1].get_legend().set_title('')
gax[1].get_legend().draw_frame(False)

# I don't want the axis with long and lat
gax[1].axis('off')

fig.subplots_adjust(hspace=0.0, wspace=0.0)
plt.show()

Print out, in a full sentence, the share of votes for Trump in a county of your choice. Express the shares with 3 decimal places.

In [19]:

#county_name = 'Clarke'
county_name = 'Gwinnett'
#county_name = 'Fulton'
tval = ga_votes.loc[(ga_votes['County']==county_name)&(ga_votes['candidate']=='Trump')&(ga_votes['category']=='Share'),'count'].mean()
bval = ga_votes.loc[(ga_votes['County']==county_name)&(ga_votes['candidate']=='Biden')&(ga_votes['category']=='Share'),'count'].mean()

# Print Results. I'm using the f-string approach to printing the results. This is the cool new way!
print(f'The share of votes attained by Donald J. Trump in {county_name} county was {tval:.3f} percent.')
print(f'The share of votes attained by Joseph R. Biden in {county_name} county was {bval:.3f} percent.')

The share of votes attained by Donald J. Trump in Gwinnett county was 40.209 percent.
The share of votes attained by Joseph R. Biden in Gwinnett county was 58.431 percent.

Practice ¶

What percent of total votes were cast before election-day, on election day, and via absentee?

In [53]:

df = ga_votes[ga_votes['category']!='Share'].groupby('category',as_index=False)['count'].sum()
df['Share'] = df['count']/df.loc[df['category']=='Total Votes','count'].values
df.set_index('category',inplace=True)

val = 100*df.loc['Advanced Voting Votes','Share']
print(f'Votes BEFORE election day: {val:.3f} percent of total.')

val = 100*df.loc['Election Day Votes','Share']
print(f'Votes ON election day: {val:.3f} percent of total.')

val = 100*df.loc['Absentee by Mail Votes','Share']
print(f'ABSENTEE votes: {val:.3f} percent of total.')

Votes BEFORE election day: 53.914 percent of total.
Votes ON election day: 19.517 percent of total.
ABSENTEE votes: 26.347 percent of total.

What share of before election-day, on election day, and via absentee voting did Biden get?

In [64]:

df = ga_votes[ga_votes['category']!='Share'].groupby(['candidate','category'],as_index=False)['count'].sum()
df['Total'] = df.groupby(['category'])['count'].transform('sum')
df['Share']= 100*df['count']/df['Total']

val = df.loc[(df['candidate']=='Biden')&(df['category']=='Advanced Voting Votes'),'Share'].values[0]
print(f'Biden vote share BEFORE election day: {val:.1f} percent of total.')

val = df.loc[(df['candidate']=='Biden')&(df['category']=='Election Day Votes'),'Share'].values[0]
print(f'Biden vote share ON election day: {val:.3f} percent of total.')

val = df.loc[(df['candidate']=='Biden')&(df['category']=='Absentee by Mail Votes'),'Share'].values[0]
print(f'Biden vote share of ABSENTEE votes: {val:.3f} percent of total.')

Biden vote share BEFORE election day: 46.4 percent of total.
Biden vote share ON election day: 37.641 percent of total.
Biden vote share of ABSENTEE votes: 64.523 percent of total.

What share of before election-day, on election day, and via absentee voting did Trump get?

In [65]:

val = df.loc[(df['candidate']=='Trump')&(df['category']=='Advanced Voting Votes'),'Share'].values[0]
print(f'Trump vote share BEFORE election day: {val:.1f} percent of total.')

val = df.loc[(df['candidate']=='Trump')&(df['category']=='Election Day Votes'),'Share'].values[0]
print(f'Trump vote share ON election day: {val:.3f} percent of total.')

val = df.loc[(df['candidate']=='Trump')&(df['category']=='Absentee by Mail Votes'),'Share'].values[0]
print(f'Trump vote share of ABSENTEE votes: {val:.3f} percent of total.')

Trump vote share BEFORE election day: 52.7 percent of total.
Trump vote share ON election day: 60.243 percent of total.
Trump vote share of ABSENTEE votes: 34.258 percent of total.

Draw a four-by-four table of chloropleth maps where colums are election-day and mail-in votes while rows correspond to Trump and Biden. Comment on the results.

In [100]:

Out[100]:

1.175394

In [105]:

# In the code below I set the maximum value for the color scale so the color schemes would be consistent across the plots.
# I also removed the legend and the quantiles argument.

fig, gax = plt.subplots(2, 2, figsize = (10,10))

maxval = ga_votes.loc[(ga_votes['category']!='Total Votes')&(ga_votes['County']!='Total:'),'count'].max()/2

#----------------------------------------------------------
# Election Day
#----------------------------------------------------------

# TRUMP MAP
ga_votes[(ga_votes['candidate']=='Trump')&(ga_votes['category']=='Election Day Votes')].plot(ax=gax[0,0], 
                                                                                             edgecolor='black', 
                                                                                             column='count',  
                                                                                             legend=False, 
                                                                                             cmap='Reds', 
                                                                                             vmin=0,
                                                                                             vmax=maxval)

val = ga_votes.loc[(ga_votes['candidate']=='Trump')&(ga_votes['category']=='Election Day Votes'),'count'].sum()/1e+6
gax[0,0].set_title(f'Trump Election Day Votes\n({val:3.2f} Million)',fontsize=18)

# gax[0,0].get_legend().set_title('')
# gax[0,0].get_legend().draw_frame(False)

# I don't want the axis with long and lat
gax[0,0].axis('off')

# BIDEN MAP
ga_votes[(ga_votes['candidate']=='Biden')&(ga_votes['category']=='Election Day Votes')].plot(ax=gax[0,1], 
                                                                                             edgecolor='black', 
                                                                                             column='count',  
                                                                                             legend=False, 
                                                                                             cmap='Blues', 
                                                                                             vmin=0,
                                                                                             vmax=maxval)

val = ga_votes.loc[(ga_votes['candidate']=='Biden')&(ga_votes['category']=='Election Day Votes'),'count'].sum()/1e+6
gax[0,1].set_title(f'Biden Election Day Votes\n({val:3.2f} Million)',fontsize=18)

# gax[0,1].get_legend().set_title('')
# gax[0,1].get_legend().draw_frame(False)

# I don't want the axis with long and lat
gax[0,1].axis('off')

#----------------------------------------------------------
# Absentee
#----------------------------------------------------------
# TRUMP MAP
ga_votes[(ga_votes['candidate']=='Trump')&(ga_votes['category']=='Absentee by Mail Votes')].plot(ax=gax[1,0], 
                                                                                                 edgecolor='black', 
                                                                                                 column='count', 
                                                                                                 legend=False, 
                                                                                                 cmap='Reds', 
                                                                                                 vmin=0,
                                                                                                 vmax=maxval)

val = ga_votes.loc[(ga_votes['candidate']=='Trump')&(ga_votes['category']=='Absentee by Mail Votes'),'count'].sum()/1e+6
gax[1,0].set_title(f'Trump Absentee Votes\n({val:3.2f} Million)',fontsize=18)

# gax[1,0].get_legend().set_title('')
# gax[1,0].get_legend().draw_frame(False)

# I don't want the axis with long and lat
gax[1,0].axis('off')

# BIDEN MAP
ga_votes[(ga_votes['candidate']=='Biden')&(ga_votes['category']=='Absentee by Mail Votes')].plot(ax=gax[1,1], 
                                                                                                 edgecolor='black', 
                                                                                                 column='count', 
                                                                                                 legend=False, 
                                                                                                 cmap='Blues', 
                                                                                                 vmin=0,
                                                                                                 vmax=maxval)

val = ga_votes.loc[(ga_votes['candidate']=='Biden')&(ga_votes['category']=='Absentee by Mail Votes'),'count'].sum()/1e+6
gax[1,1].set_title(f'Biden Absentee Votes\n({val:3.2f} Million)',fontsize=18)

# gax[1,1].get_legend().set_title('')
# gax[1,1].get_legend().draw_frame(False)

# I don't want the axis with long and lat
gax[1,1].axis('off')

# fig.subplots_adjust(hspace=0.0, wspace=0.0)
plt.show()