Time Series Analysis with Census Data

This guide provides a quick introduction to analyzing demographic changes over time using pytidycensus. For complete tutorials with detailed examples, see:

• Example 06: Simple time series with stable geographies (states, counties)

• Example 07: Advanced time series with changing boundaries (tracts, block groups)

• Example 10: Interactive 3D visualization of time series data

Installation

# For time series with automatic boundary handling
pip install pytidycensus[time]

# Basic pytidycensus (manual comparison only)
pip install pytidycensus

The Golden Rule

CRITICAL: Only compare surveys of the same type and duration:

• ACS 5-year ↔ ACS 5-year: Most stable, available for all areas

• ACS 1-year ↔ ACS 1-year: Recent trends for large areas only

• Decennial ↔ Decennial: Complete counts every 10 years

Never mix survey types (e.g., don’t compare ACS 1-year with ACS 5-year or with Decennial Census).

Quick Example 1: Simple Comparison (Stable Geographies)

For stable geographies like states or counties that don’t change boundaries, time series analysis is straightforward:

import pytidycensus as tc
import pandas as pd

# Get median household income for multiple years
income_2015 = tc.get_acs(
    geography="state",
    variables={"median_income": "B19013_001E"},
    state=["CA", "NY", "TX"],
    year=2015,
    survey="acs5"
)

income_2022 = tc.get_acs(
    geography="state",
    variables={"median_income": "B19013_001E"},
    state=["CA", "NY", "TX"],
    year=2022,
    survey="acs5"
)

# Merge and calculate change
comparison = pd.merge(
    income_2015[['NAME', 'median_income']].rename(columns={'median_income': '2015'}),
    income_2022[['NAME', 'median_income']].rename(columns={'median_income': '2022'}),
    on='NAME'
)
comparison['change'] = comparison['2022'] - comparison['2015']
comparison['pct_change'] = (comparison['change'] / comparison['2015']) * 100

See Example 06 in the examples folder for a complete tutorial on simple time series analysis.

Quick Example 2: Advanced Analysis (Changing Boundaries)

When census tract or block group boundaries change over time, pytidycensus can automatically handle this with area interpolation:

from pytidycensus.time_series import get_time_series, compare_time_periods

# Automatically interpolate data across changing boundaries
dc_data = tc.get_time_series(
    geography="tract",
    variables={"total_pop": "B01003_001E"},
    years=[2012, 2022],
    dataset="acs5",
    state="DC",
    base_year=2022,  # Use 2022 boundaries as reference
    extensive_variables=["total_pop"],
    geometry=True
)

# Compare time periods with built-in functions
comparison = tc.compare_time_periods(
    data=dc_data,
    base_period=2012,
    comparison_period=2022,
    variables=["total_pop"],
    calculate_change=True,
    calculate_percent_change=True
)

The get_time_series() function automatically:

• Detects boundary changes between years

• Performs area-weighted interpolation

• Preserves population totals

• Returns data on consistent boundaries for easy comparison

See Example 07 in the examples folder for a complete tutorial on advanced time series analysis with changing boundaries.

Quick Example 3: Changing Variable Codes

When variable codes change over time, pytidycensus can map old codes to new ones:

from pytidycensus.time_series import get_time_series
variables = {
    2010: {"total_pop": "P001001"},
    2020: {"total_pop": "P1_001N"}
}

data = get_time_series(
    geography="tract",
    variables=variables,
    years=[2010, 2020],
    dataset="decennial",
    state="DC"
)

Key Concepts

Survey Types Matter

Survey	Duration	Best For
ACS 5-year	60-month average	Small geographies, stable trends
ACS 1-year	12-month estimate	Large areas (65k+ pop), recent data
Decennial	Complete count	Decade comparisons, benchmarking

Variable Types

• Extensive variables (counts): Population, households, housing units

  • Use area-weighted sums for interpolation

• Intensive variables (rates): Percentages, densities, median income

  • Use area-weighted averages for interpolation

Boundary Changes

• States/Counties: Boundaries rarely change - simple comparison works

• Tracts/Block Groups: Boundaries change regularly - need interpolation

• ZIP codes: Not recommended for time series (unstable boundaries)

Complete Examples

For step-by-step tutorials with visualizations and detailed explanations:

• examples/06_simple_time_series_tutorial.ipynb: Basic time series with stable geographies

• examples/07_advanced_time_series_tutorial.ipynb: Advanced analysis with changing boundaries

• examples/10_interactive_3d_time_series.ipynb: Interactive 3D visualization of time series data

These tutorials cover:

• Data collection and preparation

• Handling variable code changes across years

• Visualization techniques

• Statistical analysis

• Best practices and common pitfalls