Geographic Mismatch Error: New York City District 25
This page measures two separate sources of geographic error in the estimates for New York City District 25. The first comes from fitting today’s Census tracts and block groups into the district boundary, and affects every statistic on the profile page. The second matters only for change-over-time comparisons: the Census Bureau redrew its tracts and block groups after the 2020 census, so older ACS releases were reported on different geography than current ones.
Error from Today’s Census Geography
District Geography vs. Census Geography
The map below shows how Census geographies correspond to District 25 boundaries in New York City. District boundaries are shown with a solid black outline. The selected Census geography is shown with solid gray outlines; use the layer control to switch between Census tracts and the block-group/tract hybrid. Geographies that are partially inside and partially outside a district are shaded in blue; the portion outside the district is shown with orange striped shading. Demographic data from the outside portions is borrowed into the district’s estimates and is the source of geographic mismatch error.
Map Key
Split tract — inside district
Split tract — outside district
District boundary
Tract boundary
Quantifying Geographic Aggregation Error
Using 2020 decennial census block-level data, the table below isolates the error caused by geographic aggregation. The true values come from direct aggregation of unaggregated census blocks into the district boundary. The tract and hybrid estimates are then built from the same block data after first rolling those block counts up to census tracts or block groups, and then re-estimating the district from those larger units. This shows the distortion introduced by using tract- or block-group-level geography instead of direct blocks. Because every figure comes from the same decennial census source, this comparison removes sampling noise and time mismatch and isolates the effect of aggregation alone.
| Variable | Direct Blocks (Decennial Census Data) |
Estimate (Block Data Aggregated Up)
|
Error (Estimate − True)
|
||
|---|---|---|---|---|---|
| Tract | Hybrid | Tract | Hybrid | ||
| Race | |||||
| Hispanic or Latino | 43.16 | 43.97 | 43.39 | +0.82 pp | +0.24 pp |
| White | 10.96 | 10.94 | 10.85 | -0.02 pp | -0.11 pp |
| Black | 1.51 | 1.56 | 1.56 | +0.05 pp | +0.04 pp |
| Asian | 41.61 | 40.78 | 41.42 | -0.84 pp | -0.19 pp |
| Other or Multi-racial | 2.76 | 2.75 | 2.78 | +0.00 pp | +0.02 pp |
| Average Absolute Error | — | — | — | 0.34 pp | 0.12 pp |
| Age Group | |||||
| Under 5 years | 4.78 | 4.80 | 4.77 | +0.01 pp | -0.01 pp |
| 5 to 17 years | 13.58 | 13.65 | 13.63 | +0.07 pp | +0.05 pp |
| 18 to 24 years | 8.18 | 8.21 | 8.21 | +0.04 pp | +0.03 pp |
| 25 to 44 years | 31.87 | 32.07 | 31.97 | +0.21 pp | +0.11 pp |
| 45 to 64 years | 26.98 | 26.85 | 26.86 | -0.14 pp | -0.12 pp |
| 65 years and over | 14.61 | 14.42 | 14.55 | -0.19 pp | -0.06 pp |
| Average Absolute Error | — | — | — | 0.11 pp | 0.06 pp |
| Housing Tenure | |||||
| Owner-occupied | 32.92 | 32.06 | 32.68 | -0.86 pp | -0.24 pp |
| Renter-occupied | 67.08 | 67.94 | 67.32 | +0.86 pp | +0.24 pp |
| Average Absolute Error | — | — | — | 0.86 pp | 0.24 pp |
| Housing Occupancy | |||||
| Occupied Housing | 95.21 | 95.30 | 95.28 | +0.09 pp | +0.07 pp |
| Vacant Housing | 4.79 | 4.70 | 4.72 | -0.09 pp | -0.07 pp |
| Average Absolute Error | — | — | — | 0.09 pp | 0.07 pp |
| Household Size | |||||
| 1-person HH | 22.95 | 22.85 | 22.86 | -0.10 pp | -0.09 pp |
| 2-person HH | 25.61 | 25.57 | 25.66 | -0.03 pp | +0.05 pp |
| 3-person HH | 18.63 | 18.60 | 18.59 | -0.03 pp | -0.04 pp |
| 4+ person HH | 32.81 | 32.97 | 32.88 | +0.16 pp | +0.07 pp |
| Average Absolute Error | — | — | — | 0.08 pp | 0.06 pp |
| Household Type | |||||
| Family Households | 69.16 | 69.16 | 69.18 | +0.00 pp | +0.02 pp |
| Married-couple Family | 44.15 | 44.08 | 44.15 | -0.07 pp | +0.00 pp |
| Non-family Households | 30.84 | 30.84 | 30.82 | +0.00 pp | -0.02 pp |
| Living Alone | 22.95 | 22.85 | 22.86 | -0.10 pp | -0.09 pp |
| Average Absolute Error | — | — | — | 0.04 pp | 0.03 pp |
| Average Absolute Error Across Categories | — | — | — | 0.26 pp | 0.10 pp |
| Note: Percentages may not sum to 100% due to rounding. | |||||
| Average Absolute Error: mean of absolute errors across variables within each category. | |||||
| Average Absolute Error Across Categories: simple average of the six category-level averages. | |||||
Error in Change Over Time
The section above quantifies the error from squeezing today’s Census geography into district boundaries. This section quantifies a second, separate source of error that only matters for change-over-time comparisons: the Census Bureau redraws its tracts and block groups after each decennial census. All ACS releases from 2010 through 2019 report data on 2010-vintage geography, while current releases use 2020-vintage geography. When we compare a district’s statistics across those eras, part of any apparent “change” can come purely from the redrawn Census geography rather than from real demographic change.
To isolate that effect, we hold the demographic data completely fixed — the same 2020 decennial census block counts — and estimate New York City District 25 twice: once through current 2020-vintage geography, and once through the 2010-vintage geography used by older ACS releases. Any difference between the two estimates is caused by the geography vintage alone.
District Geography: Today vs. 2010 Census Geography
The map below shows how Census geographies correspond to the New York City District 25 boundary in each geography vintage. Use the layer control to switch between Census tracts and block groups, in both the current 2020 vintage and the 2010 vintage used by older ACS releases. Units shaded in blue were redrawn between the two vintages — typically a 2010 unit that was split into multiple 2020 units — and are relevant to the district boundary in at least one vintage. These redrawn units are where the geography vintage can move the district’s estimates.
Map Key
Redrawn between 2010 and 2020 vintages
District boundary
Tract / block group boundary (selected layer)
Quantifying Historical Geography Error
The table below isolates how much the geography vintage alone moves New York City District 25’s statistics, at both the tract and block-group level. Every column is computed from the same 2020 decennial census block data, rolled up once using current 2020-vintage geography and once using 2010-vintage geography, and then re-estimated for the district from those larger units. The vintage difference is the current-geography estimate minus the 2010-geography estimate: it is the portion of an apparent over-time “change” that would be created by the redrawn geography even if no person had moved.
| Variable |
Tract Estimate (Same 2020 Data)
|
Block Group Estimate (Same 2020 Data)
|
||||
|---|---|---|---|---|---|---|
| 2020 Geography |
2010 Geography |
Diff | 2020 Geography |
2010 Geography |
Diff | |
| Race | ||||||
| Hispanic or Latino | 43.97 | 44.31 | -0.34 pp | 43.39 | 43.25 | +0.14 pp |
| White | 10.94 | 10.84 | +0.10 pp | 10.85 | 10.78 | +0.07 pp |
| Black | 1.56 | 1.60 | -0.04 pp | 1.56 | 1.60 | -0.04 pp |
| Asian | 40.78 | 40.52 | +0.26 pp | 41.42 | 41.60 | -0.18 pp |
| Other or Multi-racial | 2.75 | 2.73 | +0.02 pp | 2.78 | 2.77 | +0.00 pp |
| Average Absolute Difference | — | — | 0.15 pp | — | — | 0.09 pp |
| Age Group | ||||||
| Under 5 years | 4.80 | 4.81 | -0.01 pp | 4.77 | 4.78 | -0.01 pp |
| 5 to 17 years | 13.65 | 13.60 | +0.05 pp | 13.63 | 13.55 | +0.08 pp |
| 18 to 24 years | 8.21 | 8.21 | +0.00 pp | 8.21 | 8.19 | +0.02 pp |
| 25 to 44 years | 32.08 | 31.99 | +0.09 pp | 31.98 | 31.91 | +0.07 pp |
| 45 to 64 years | 26.85 | 26.82 | +0.03 pp | 26.86 | 26.86 | +0.00 pp |
| 65 years and over | 14.42 | 14.58 | -0.16 pp | 14.55 | 14.72 | -0.17 pp |
| Average Absolute Difference | — | — | 0.06 pp | — | — | 0.06 pp |
| Housing Tenure | ||||||
| Owner-occupied | 32.06 | 32.28 | -0.23 pp | 32.68 | 33.00 | -0.32 pp |
| Renter-occupied | 67.94 | 67.72 | +0.23 pp | 67.32 | 67.00 | +0.32 pp |
| Average Absolute Difference | — | — | 0.23 pp | — | — | 0.32 pp |
| Housing Occupancy | ||||||
| Occupied Housing | 95.30 | 95.37 | -0.07 pp | 95.28 | 95.32 | -0.04 pp |
| Vacant Housing | 4.70 | 4.63 | +0.07 pp | 4.72 | 4.68 | +0.04 pp |
| Average Absolute Difference | — | — | 0.07 pp | — | — | 0.04 pp |
| Household Size | ||||||
| 1-person HH | 22.85 | 22.79 | +0.06 pp | 22.86 | 22.91 | -0.05 pp |
| 2-person HH | 25.57 | 25.70 | -0.13 pp | 25.66 | 25.83 | -0.17 pp |
| 3-person HH | 18.60 | 18.59 | +0.01 pp | 18.59 | 18.57 | +0.02 pp |
| 4+ person HH | 32.97 | 32.91 | +0.06 pp | 32.88 | 32.69 | +0.19 pp |
| Average Absolute Difference | — | — | 0.06 pp | — | — | 0.11 pp |
| Household Type | ||||||
| Family Households | 69.16 | 69.29 | -0.13 pp | 69.18 | 69.19 | -0.01 pp |
| Married-couple Family | 44.08 | 44.12 | -0.04 pp | 44.15 | 44.15 | +0.01 pp |
| Non-family Households | 30.84 | 30.71 | +0.13 pp | 30.82 | 30.81 | +0.01 pp |
| Living Alone | 22.85 | 22.79 | +0.06 pp | 22.86 | 22.91 | -0.05 pp |
| Average Absolute Difference | — | — | 0.09 pp | — | — | 0.02 pp |
| Average Absolute Difference Across Categories | — | — | 0.11 pp | — | — | 0.11 pp |
| Note: Percentages may not sum to 100% due to rounding. | ||||||
| Average Absolute Difference: mean of absolute differences across variables within each category. | ||||||
| Average Absolute Difference Across Categories: simple average of the six category-level averages. | ||||||
How to Read These Numbers
This is a geography-vintage sensitivity test, not a measure of demographic change. Because the underlying data never changes — every estimate above uses the same 2020 census block counts — the vintage differences show how much of an over-time comparison could be an artifact of redrawn Census geography rather than real change. For New York City District 25’s tract estimate, the largest single-variable effect is -0.34 pp (Hispanic or Latino share), and the average absolute effect across categories is 0.11 pp.
A few caveats:
- The 2010 vintage covers all pre-2020 ACS releases. Census tract and block group boundaries are held fixed between decennial censuses, so every ACS window from 2010 through 2019 (including 2015–2019) is tabulated on the same 2010-vintage geography compared here. The mapping between vintages uses the NHGIS 2020-block-to-2010-block crosswalk.
- Large vintage effects are a warning, not a correction. They flag which statistics in this district are geography-sensitive when comparing older and newer ACS windows; they do not measure what actually changed.
- The published change-over-time series on the profile pages crosswalks each historical ACS window to current geography before aggregating, which removes most — but not all — of this vintage sensitivity. This section bounds how large the residual geography effect could be.