Census estimates offer a clearer picture of renter cost burden when the survey period, household definitions, and uncertainty stay visible.
Data snapshot: September 18, 2026. Census coverage: 2020 through 2024.
A monthly rent figure tells only part of a household’s housing story. The same payment can leave one household with considerable flexibility and another with very little room for food, transport, medical expenses, or savings. To understand housing pressure, it helps to examine the relationship between rent and household income rather than treating the rent amount as a complete measure of affordability.
For San Francisco County, the 2024 American Community Survey estimates covering five years identify 224,913 rental housing units whose occupants pay cash rent. Within that total, an estimated 42,422 units were occupied by households whose gross rent represented at least half of household income. These are survey estimates for the period from 2020 through 2024, not a count of today’s tenants or a finding from Homzora’s own renter survey. Read the published Census extract.
The distinction matters. A carefully described Census estimate can reveal the scale and distribution of housing pressure. A loosely described estimate can create the impression that a historical survey provides an exact reading of the current market. This article explains what the available figures show, how the calculations work, and where the evidence stops.
Start with the population being measured
The source is Census table B25070, which describes gross rent as a percentage of household income. Its universe is rental housing units whose occupants pay cash rent. That wording is more specific than all San Francisco residents, all renters, or all households. Homes occupied by their owners are outside the table, and rental arrangements without cash rent are not included in this universe.
The unit of measurement is also important. A housing unit occupied by several people does not become several separate observations merely because more than one person lives there. These counts therefore should not be described as the number of individual renters experiencing a particular burden. Using the correct unit keeps an apparently simple statistic from becoming misleading.
The geography is San Francisco County, identified in the dataset by Census geographic code 06075. The figures should not be labeled as estimates for the entire Bay Area. They also do not provide separate findings for the neighborhood where a reader happens to live. The citywide picture is informative, but it does not replace a more specific local analysis.
Gross rent is different from advertised rent
Gross rent includes contract rent together with applicable utility and fuel costs. It is designed to describe a broader housing expense than the rent number printed at the top of an advertisement. That definition helps explain why the measure should not be compared casually with a listing price that excludes some recurring expenses. The source variables and estimates appear in the Census Bureau’s B25070 data file.
For a simple illustration, imagine a household paying $2,800 in contract rent and another $200 in applicable utilities. The combined amount is $3,000. If household income is $10,000 per month, that illustrative housing amount equals 30 percent of income. Looking only at the contract rent would produce a different percentage.
This example is arithmetic, not an observation about a particular San Francisco household. Its purpose is to show why definitions matter. Before comparing a household budget with a published statistic, a reader should identify which expenses are included in each number and whether the income measure is being used consistently.
What the Census distribution shows
The table below reproduces the estimated counts for the higher rent burden categories, along with the total universe and the category for which a percentage was not computed. Each margin of error is reported at the 90 percent confidence level. The total and component rows overlap, so they must not be added together.
| Gross rent as a share of household income | Estimated housing units | 90 percent margin of error |
|---|---|---|
| Total universe | 224,913 | ±2,332 |
| 30.0 to 34.9 percent | 17,016 | ±1,202 |
| 35.0 to 39.9 percent | 10,830 | ±941 |
| 40.0 to 49.9 percent | 14,732 | ±1,036 |
| 50.0 percent or more | 42,422 | ±1,631 |
| Percentage not computed | 10,344 | ±978 |
Source: U.S. Census Bureau, 2024 ACS estimates covering five years, B25070. Download all categories and margins of error.
The largest of the four displayed burden categories is the group spending at least half of household income on gross rent. That observation concerns the published estimates in this table. It does not establish why a household falls into that category, how long it has been there, or whether its circumstances have changed since the survey period.
How to calculate a share without hiding the denominator
Adding the four categories beginning at 30 percent produces an estimated 85,000 housing units. Subtracting the 10,344 units in the percentage not computed category from the total leaves an estimated 214,569 units with a computed percentage of income spent on rent. Dividing 85,000 by 214,569 produces approximately 39.6 percent.
That is a Homzora calculation from the published category estimates. Its precise interpretation is that approximately 39.6 percent of the units with a computed percentage fell into the categories spending at least 30 percent of household income on gross rent. It should not be shortened to a claim that 39.6 percent of all San Francisco residents currently struggle to pay rent.
Using the same denominator, the category spending at least half of income represents approximately 19.8 percent. These calculated shares describe point estimates only. This article does not calculate margins of error for the combined numerator or resulting percentages, and it makes no statistical significance claim. The source margins of error cannot simply be added to produce a valid margin for a combined result.
The denominator is not a technical detail that can be omitted without consequence. Including units whose percentage was not computed would answer a different question. Treating those units as having no housing burden would introduce an unsupported assumption. Stating the denominator allows readers to understand exactly which observations contribute to the calculation.
Why the survey period must remain visible
The 2024 release covering five years combines information collected across 2020 through 2024. It is a period estimate rather than a snapshot of one month or a survey conducted only during 2024. The Census Bureau explains the differences between its products covering one year and five years in its guidance on choosing ACS estimates.
For someone reading this article in September 2026, that means the figures describe an earlier period. They remain useful for understanding the distribution captured by that survey product. They do not establish how the distribution has changed during 2025 or 2026, and they should not be presented as proof that every household’s current burden matches its earlier circumstances.
The timing issue also affects comparisons with other datasets. An August 2026 market rent observation and an estimate covering 2020 through 2024 do not describe the same moment. Placing them next to each other can provide context, but it does not produce a valid calculation of the current burden faced by a typical household.
Income and housing costs can change independently
A household’s rent burden can increase even when its rent stays the same. A reduction in working hours, the departure of an earning household member, or another change in income can alter the percentage. Conversely, a household may face a higher rent while spending a smaller share of income if its income rises by a larger proportion.
Consider two hypothetical households paying the same $3,000 monthly housing amount. One has $10,000 in monthly income, while the other has $6,000. The corresponding shares are 30 percent and 50 percent. Nothing about that arithmetic requires the apartments to differ. The income denominator changes the relationship.
The Census distribution does not identify the cause of every household’s position within it. It cannot show that high rent alone explains each higher burden category, nor can it tell readers which households recently experienced income changes. These limitations are a reason to ask more precise questions, not a reason to disregard the evidence.
A percentage does not describe the entire household budget
Two households with the same housing percentage may have very different amounts of money remaining after housing costs. They may also face different necessary expenses. Household size, transport requirements, medical needs, care responsibilities, and existing obligations can all affect how much flexibility remains.
For that reason, a percentage of income spent on rent is a useful descriptive measure rather than a complete assessment of personal financial security. The table does not measure savings, emergency reserves, debt balances, or the quality of the housing itself. A reader should resist treating one percentage as a complete judgment about a household’s circumstances.
Homzora’s San Francisco housing budget guide can help readers organize a separate personal comparison. The aim is to identify recurring housing expenses, other necessary spending, and upfront cash needs in a consistent format. That exercise complements a population statistic without pretending that the statistic has supplied a personalized recommendation.
What margins of error add to the discussion
The published counts are estimates rather than exact administrative totals. The accompanying margins of error communicate sampling uncertainty. Keeping them beside the figures helps readers recognize that a reported value such as 42,422 should not be interpreted as a perfectly precise census of every relevant housing unit.
A margin of error also does not capture every possible limitation of a dataset. It does not make an older period current, repair a mismatch in definitions, or justify a comparison between different geographic areas. Those issues need their own checks even when the statistical uncertainty is relatively small.
When future articles compare periods or places, they should examine whether the underlying measures are comparable and whether the apparent difference supports the proposed conclusion. This article does not perform such a test. It explains one published distribution and a small number of transparent calculations based on that distribution.
Keep Census findings separate from original renter research
Homzora’s San Francisco renter survey is a separate research project. Its voluntary responses should not be combined with Census estimates as though the two sources came from the same sample. They have different collection methods, coverage, and intended uses.
No San Francisco survey findings are being published through this article. The existing research methodology requires sufficient eligible responses, a completed quarter, and editorial review before proposed measures can be released. Meeting a minimum response count would not, by itself, make a voluntary sample representative of every renter in the city.
The Census figures provide a documented baseline for understanding the housing costs and incomes captured during their survey period. Original renter research may eventually add another perspective, provided its limitations remain explicit. Readers can explore the source estimates, definitions, and related measures in Homzora’s San Francisco data library and use them to ask better questions about the housing situations they encounter.