Population Change 2020-2021

ACS Data

Part 1 of this notebook demonstrates how to query and merge the same data set from different years. It also demonstrates how, behind the scenes, censusdis loads metadata and works across multiple years. Nothing is hard-coded in the library about what variables or data sets are available in what years. So when the 2021 data was published, this notebook just worked.

However, it also demonstrates some issues one must be keenly aware of when looking at ACS data. ACS5 data covers a wide range of variables, but it also computes representative numbers over a five year time period, not numbers as of a specific data as some other data sets do. So the values presented in the 2020 data set are based on survey results from 2016, 2017, 2018, 2019, and 2020. If we compare 2021 ACS5 to 2020 ACS5, 4/5 of the sample comes from 2017, 2018, 2019 and 2020 in both. In the 2020 data set, there is also data from 2016, whereas in the 2021 data set there is also data from 2021.

PEP Data

Part 2 looks at the Population Estimates (PEP) data set, which seeks to estimate population at a specific point in time, specifically June 1 of each year. It also tracks changes from the previous year in both numeric and percentage terms.

As we can see by comparing the data and maps in parts 1 and 2, it is important to understand not just what was measured in each data set, but how. Look at the difference in New York State for a particularly extreme difference based on which data set we look at.

# So we can run from within the censusdis project and find the packages we need.
import os
import sys

sys.path.append(
    os.path.join(os.path.abspath(os.path.join(os.path.curdir, os.path.pardir)))
)

Imports and configuration

import censusdis.data as ced
import censusdis.maps as cem

from censusdis.states import STATE_NY

Part 1: Using ACS5 Data

DATASET_ACS5 = "acs/acs5"

TOTAL_POPULATION_VARIABLE = "B01003_001E"

gdf_states_2020 = ced.download(
    DATASET_ACS5,
    2020,
    ["NAME", TOTAL_POPULATION_VARIABLE],
    state="*",
    with_geometry=True,
)

# Note that we don't load the geometry here.
# We already have it above, and it will
# still be available after the merge.
df_states_2021 = ced.download(
    DATASET_ACS5, 2021, ["NAME", TOTAL_POPULATION_VARIABLE], state="*"
)

gdf_both_years = gdf_states_2020.merge(
    df_states_2021, on=["STATE", "NAME"], suffixes=("_2020", "_2021")
)

TOTAL_POPULATION_2020_VARIABLE = f"{TOTAL_POPULATION_VARIABLE}_2020"
TOTAL_POPULATION_2021_VARIABLE = f"{TOTAL_POPULATION_VARIABLE}_2021"

TOTAL_POPULATION_CHANGE = f"{TOTAL_POPULATION_VARIABLE}_DELTA"
TOTAL_POPULATION_CHANGE_PERCENT = f"{TOTAL_POPULATION_CHANGE}_PERCENT"

gdf_both_years[TOTAL_POPULATION_CHANGE] = (
    gdf_both_years[TOTAL_POPULATION_2021_VARIABLE]
    - gdf_both_years[TOTAL_POPULATION_2020_VARIABLE]
)

gdf_both_years[TOTAL_POPULATION_CHANGE_PERCENT] = (
    100
    * gdf_both_years[TOTAL_POPULATION_CHANGE]
    / gdf_both_years[TOTAL_POPULATION_2020_VARIABLE]
)

Absolute

gdf_both_years.sort_values(TOTAL_POPULATION_CHANGE, ascending=False)[
    ["NAME", TOTAL_POPULATION_CHANGE]
].head()

	NAME	B01003_001E_DELTA
4	New York	599896
29	New Jersey	348606
24	Texas	227139
0	Pennsylvania	175765
7	Florida	122838

Percentagewise

gdf_both_years.sort_values(TOTAL_POPULATION_CHANGE_PERCENT, ascending=False)[
    ["NAME", TOTAL_POPULATION_CHANGE_PERCENT]
].head()

	NAME	B01003_001E_DELTA_PERCENT
29	New Jersey	3.923349
19	Idaho	3.263285
42	Rhode Island	3.228499
4	New York	3.074049
32	Vermont	2.770446

Fastest shrinking states

Absolute

gdf_both_years.sort_values(TOTAL_POPULATION_CHANGE, ascending=True)[
    ["NAME", TOTAL_POPULATION_CHANGE]
].head()

	NAME	B01003_001E_DELTA
16	Arizona	-94861
33	North Carolina	-19205
5	District of Columbia	-18820
39	Mississippi	-14812
8	South Carolina	-12614

Percentagewise

gdf_both_years.sort_values(TOTAL_POPULATION_CHANGE_PERCENT, ascending=True)[
    ["NAME", TOTAL_POPULATION_CHANGE_PERCENT]
].head()

	NAME	B01003_001E_DELTA_PERCENT
5	District of Columbia	-2.681011
16	Arizona	-1.322277
15	Wyoming	-0.809670
39	Mississippi	-0.496741
2	West Virginia	-0.352822

Plot on maps

from matplotlib.ticker import FuncFormatter

ax = cem.plot_us(
    gdf_both_years,
    TOTAL_POPULATION_CHANGE,
    legend=True,
    figsize=(12, 6),
    cmap="PRGn",
    edgecolor="black",
    linewidth=0.25,
    vmin=-600_000,
    vmax=600_000,
    legend_kwds={"format": FuncFormatter(lambda x, pos: f"{x:+,.0f}")},
)

ax.axis("off")
ax.set_title("Total Population Change by State, 2020-2021")

Text(0.5, 1.0, 'Total Population Change by State, 2020-2021')

ax = cem.plot_us(
    gdf_both_years,
    TOTAL_POPULATION_CHANGE_PERCENT,
    legend=True,
    figsize=(12, 6),
    cmap="PRGn",
    edgecolor="black",
    linewidth=0.25,
    vmin=-4,
    vmax=4,
    legend_kwds={"format": FuncFormatter(lambda x, pos: f"{x:+.1f}%")},
)

ax.axis("off")
ax.set_title("Total Population Percentage Change by State, 2020-2021")

Text(0.5, 1.0, 'Total Population Percentage Change by State, 2020-2021')

Part 2: Using the Population Estimates Dataset

Instead of using two different years of the ACS5 data set, we could also use a specialized U.S. Census Data Set that estimates population and population change. It is called pep/population and is described at https://api.census.gov/data/2021/pep/population.html.

DATASET_PEP = "pep/population"

# See https://api.census.gov/data/2021/pep/population/variables.html

POP_2020_VARIABLE = "POP_2020"
POP_2021_VARIABLE = "POP_2021"
NUM_CHANGE_VARIABLE = "NPOPCHG_2021"
PCT_CHANGE_VARIABLE = "PPOPCHG_2021"

PEP_VARIABLES = [
    "NAME",
    POP_2020_VARIABLE,
    POP_2021_VARIABLE,
    NUM_CHANGE_VARIABLE,
    PCT_CHANGE_VARIABLE,
]

gdf_pep = ced.download(DATASET_PEP, 2021, PEP_VARIABLES, state="*", with_geometry=True)

Absolute

gdf_pep[["NAME", NUM_CHANGE_VARIABLE]].sort_values(
    NUM_CHANGE_VARIABLE, ascending=False
).head()

	NAME	NPOPCHG_2021
10	Texas	310288
40	Florida	211196
45	Arizona	98330
15	North Carolina	93985
25	Georgia	73766

Percentagewise

gdf_pep[["NAME", PCT_CHANGE_VARIABLE]].sort_values(
    PCT_CHANGE_VARIABLE, ascending=False
).head()

	NAME	PPOPCHG_2021
34	Idaho	2.876491
46	Utah	1.715308
36	Montana	1.664345
45	Arizona	1.369883
17	South Carolina	1.168957

Fastest shrinking states

Absolute

gdf_pep[["NAME", NUM_CHANGE_VARIABLE]].sort_values(
    NUM_CHANGE_VARIABLE, ascending=True
).head()

	NAME	NPOPCHG_2021
37	New York	-319020
19	California	-261902
43	Illinois	-113776
42	Massachusetts	-37497
21	Louisiana	-27156

Percentagewise

gdf_pep[["NAME", PCT_CHANGE_VARIABLE]].sort_values(
    PCT_CHANGE_VARIABLE, ascending=True
).head()

	NAME	PPOPCHG_2021
9	District of Columbia	-2.904391
37	New York	-1.582838
43	Illinois	-0.889901
2	Hawaii	-0.713405
19	California	-0.663047

Plot on maps

from matplotlib.ticker import FuncFormatter

ax = cem.plot_us(
    gdf_pep,
    NUM_CHANGE_VARIABLE,
    legend=True,
    figsize=(12, 6),
    cmap="PRGn",
    edgecolor="black",
    linewidth=0.25,
    vmin=-600_000,
    vmax=600_000,
    legend_kwds={"format": FuncFormatter(lambda x, pos: f"{x:+,.0f}")},
)

ax.axis("off")
ax.set_title("PEP Data - Total Population Change by State, 2020-2021")

Text(0.5, 1.0, 'PEP Data - Total Population Change by State, 2020-2021')

ax = cem.plot_us(
    gdf_pep,
    PCT_CHANGE_VARIABLE,
    legend=True,
    figsize=(12, 6),
    cmap="PRGn",
    edgecolor="black",
    linewidth=0.25,
    vmin=-4,
    vmax=4,
    legend_kwds={"format": FuncFormatter(lambda x, pos: f"{x:+.1f}%")},
)

ax.axis("off")
ax.set_title("PEP Data - Total Population Percentage Change by State, 2020-2021")

Text(0.5, 1.0, 'PEP Data - Total Population Percentage Change by State, 2020-2021')

Part 3. Compare NY Results

gdf_both_years[gdf_both_years.STATE == STATE_NY][
    ["NAME", TOTAL_POPULATION_CHANGE, TOTAL_POPULATION_CHANGE_PERCENT]
]

	NAME	B01003_001E_DELTA	B01003_001E_DELTA_PERCENT
4	New York	599896	3.074049

gdf_pep[gdf_pep.STATE == STATE_NY][["NAME", NUM_CHANGE_VARIABLE, PCT_CHANGE_VARIABLE]]

	NAME	NPOPCHG_2021	PPOPCHG_2021
37	New York	-319020	-1.582838