Methodology

Data Source: LODES

CensusFlow uses the LEHD Origin-Destination Employment Statistics (LODES) dataset, version 8, published by the U.S. Census Bureau as part of the Longitudinal Employer-Household Dynamics (LEHD) program.

LODES provides workplace and residential location data at the Census block level for nearly all jobs covered by state unemployment insurance programs. The data undergoes noise infusion by the Census Bureau to protect individual privacy while preserving aggregate statistical accuracy.

LODES File Types

We use three core LODES file types to build each commute page:

Earnings Tiers

LODES splits workers into three earnings categories based on monthly earnings:

CensusFlow calculates state-level tertile thresholds from the aggregate distribution of workers across these categories, then applies them as cutpoints for the earnings color scale on each map.

Geographic Aggregation

Census Places (Cities)

LODES data is recorded at the Census block level. We aggregate blocks up to Census Places using the LODES crosswalk file, which maps each block to its enclosing state, county, Census tract, and place. City pages on CensusFlow correspond to Census-designated places — incorporated cities, towns, CDPs, and boroughs.

ZIP Code Tabulation Areas (ZCTAs)

ZIP pages use Census ZCTAs, which are generalized area representations of USPS ZIP code service areas. ZCTAs are not identical to ZIP codes but serve as the best available geographic proxy for them in Census data.

Commute Distance

Average commute distance is calculated using the haversine formula between the population-weighted centroids of the origin and destination geographies (Census Place or ZCTA). This gives a straight-line (“as the crow flies”) distance, which underestimates actual road travel distance but provides a consistent, comparable metric across all geographies.

Work-From-Home Proxy

LODES records where a worker lives and where they work. When both the home and work location fall within the same Census block, we flag this as a “same-block” match. Because Census blocks in residential areas are typically small, a same-block home-work match strongly suggests the worker is self-employed or working from home.

This is a proxy measure — not all same-block matches are remote workers (e.g., a worker may live above their shop), and not all remote workers will have same-block matches. However, it provides a useful lower-bound estimate of work-from-home rates derived purely from administrative records.

Dot-Density Map Rendering

The interactive maps use a dot-density approach where each dot represents approximately 75 workers. Dots are randomly jittered within their origin or destination geography to prevent overlap and create a natural density gradient. The map renders using MapLibre GL JS and deck.gl for GPU-accelerated WebGL visualization.

Flow lines in “Streams” mode connect origin and destination centroids with curved arcs, colored by the dominant earnings tier of the flow.