A count of records in San Antonio's permit portal is not automatically a count of new homes authorized by the Census Building Permits Survey. The city data can include applications, issued permits and trade work, while the Census series has a defined residential construction scope. Before charting a trend, decide which question the data should answer and identify the unit being counted. Otherwise a spreadsheet can produce a clean trend from incompatible records.
San Antonio's Development Services reports page links separate application and issued permit datasets through Open Data SA. Census publishes Building Permits Survey data at several geographic levels and describes new privately owned residential construction. This guide explains how to prepare a reconciliation worksheet between those sources. It does not report a newly collected five year series or invent local totals. The worked records are fictional and illustrate why definitions must be settled before counting.
Name the research question precisely
A question about new housing authorizations differs from a question about all building activity or the number of applications submitted. Write the intended measure in a sentence before downloading data. For example, you may want the number of housing units authorized in a defined period and geography. That wording identifies a unit, event and scope. A vague request for permits in San Antonio leaves all three open to interpretation.
Keep the question visible at the top of the analysis file. When a field or source does not match it, decide whether to change the source or narrow the question, and record the decision. Do not quietly switch from housing units to permit records because the latter is easier to count. A smaller but well defined analysis is more useful than a larger chart whose title promises something the data do not measure.
Distinguish submitted applications from issued permits
The city's reports page explicitly separates building permit applications submitted from permits issued. Select the dataset corresponding to the event you intend to study and record its title. An application can be part of an administrative pipeline without being an authorization. Combining application and issuance rows may count the same project more than once or mix different stages under one total.
If you need both stages, analyze them separately and use identifiers to study relationships where the data support that work. Do not add the two counts and label the sum new housing supply. The comparison should show what each event means and which date field was used. A submission date and issue date can fall in different years even for the same project, so period selection must follow the chosen measure.
Read Census's unit definitions
Census defines authorization in relation to local approval through building or zoning permits and classifies residential structures by their housing unit characteristics. The Building Permits Survey concerns new privately owned residential construction, not every repair or trade permit in a city portal. Record whether a Census column counts buildings, units or another quantity. A multifamily structure can contain multiple housing units, making those measures materially different.
Keep structure categories in their original form until you understand the table. A one unit structure category is not simply every record whose description contains residential. Likewise, several permits connected to one development do not necessarily represent several separate buildings. The local data's record structure and the Census definitions must be read together before deciding how to aggregate a city export for comparison.
Match the geography without relying on a title
Census makes data available at national, state, metropolitan, county and permit issuing place levels. San Antonio city and the San Antonio metropolitan area are different scopes. Record the exact geography and identifier from the file used. Do not select a metropolitan series because its title begins with San Antonio and then describe the result as construction solely inside the city.
The city portal's coverage should also be documented. A permit issuing jurisdiction can have a different relationship to statistical boundaries than a casual map label suggests. Census's methodology discusses permit issuing places and avoiding duplicate reporting. If you cannot establish that the local export and Census geography align, present them as separate sources with different scopes rather than forcing a reconciliation that implies they should have identical totals.
Save the download and its metadata
Keep the original file, access date, dataset title, filters and field definitions. If you use Open Data SA's interface to filter before downloading, record those selections. The city's reports page describes filtering and downloading through its reporting platform, but a saved CSV alone may not reveal every filter applied. A reproducible analysis needs both the data and the method used to obtain it.
For Census, record the release and whether the figures are monthly, year to date or final annual data. The survey publishes multiple frequencies and revisions. Do not compare a preliminary monthly sum with a later final annual figure without explaining the difference in vintage. The same calendar year label does not guarantee that two downloaded files contain the same version of the estimates.
A fictional project with several records
Imagine a fictional development with one building containing twenty homes. The city export contains an application record, an issued building permit and several related trade permits. Counting every row would not yield twenty housing units, and counting only the building permit would yield one record rather than twenty units. The appropriate housing unit measure requires a supported unit field or other documented source, not a guess based on the number of rows.
The researcher creates separate columns for record identifier, event type, project relationship and housing units where documented. Missing unit information remains missing. The researcher does not assign one unit to every blank row. This fictional example shows why a permit portal can contain valuable project evidence while still requiring careful transformation before it can answer a housing authorization question.
Define exclusions before calculating totals
Read the local fields and decide how to distinguish new residential construction from alterations, repairs and nonresidential work. Document the exact criteria and retain ambiguous records for review. Do not use a broad keyword filter without checking what it includes and excludes. A description mentioning apartments might concern a repair rather than new homes, while a new project may use an abbreviated category that a keyword search misses.
Create an exclusion log with reasons such as wrong event, wrong geography, nonresidential scope or unresolved classification. This does not need to expose personal applicant information. The log's purpose is methodological transparency. If another researcher disagrees with a rule, they should be able to see which records would change, rather than trying to reverse engineer a total from an undocumented spreadsheet.
Handle duplicates and revisions explicitly
Identify the key that represents a unique record and the fields that connect revisions or related permits. Do not delete similar rows merely because their addresses match. They may represent different authorized work. Conversely, repeated exports or updated versions of one record should not automatically count as separate projects. Use the dataset's documentation and agency clarification where necessary to establish the relationship.
When a Census release revises earlier figures, retain the old file and label the newer vintage. A revision does not necessarily mean the earlier analyst made an arithmetic mistake. It can reflect the survey's update process. A trend chart should use a consistent version where possible and state the release used. Avoid mixing old and new values selectively because one combination produces a more dramatic story.
Build a comparison table before a chart
The table should include year or month, exact geography, source, event, unit counted, release date and value. Add a note for missing or unresolved information. A blank is not automatically zero. If the data use a special missing code, preserve its meaning in the analysis rather than converting it to a numeric value for convenience. The chart can be created only after these definitions are stable.
For a multiyear comparison, use the same criteria across periods or clearly explain a documented change. A new portal or field definition can create an apparent trend unrelated to construction. Check whether the available record system and coverage changed. If consistency cannot be established, limit the comparison to a defensible period instead of filling a five year chart with incompatible data merely to satisfy a visual format.
Keep authorizations separate from completed homes
A permit authorization is not a completion date, an occupied unit or a currently available rental. Do not use the count alone to promise new supply by a particular month or predict rent changes. Those questions require additional evidence and a method connecting stages. A local project record may supply some later information, but it should be matched carefully rather than assumed from the passage of time.
Likewise, a larger authorization total does not by itself explain why it changed. Project timing, structure mix, reporting and other factors may matter. The data can describe the measured change under the chosen definitions. A causal explanation requires more work. Use language such as authorized units recorded rather than homes delivered when the source measures authorization.
The finished San Antonio research file should show the question, original sources, geography, event and counting rules before presenting any total. The city data can illuminate local project records, while Census provides a defined statistical series. They are most useful when their differences are made explicit. A transparent reconciliation method produces a stronger foundation for a later chart than an immediate row count labeled housing growth.