Chapter 1. Start with the question, not the number A housing number is useful only when it answers the question in front of you. A county rent estimate can tell you something about the housing costs paid by a broad group of households. It cannot tell you the right asking rent for a particular apartment. A property tax record can tell you how the appraiser treated a parcel for a particular tax year. It cannot promise the tax bill after a purchase. A flood map can tell you the mapped hazard classification under a particular map. It cannot tell you that a house will stay dry. These distinctions sound small until money or a deadline depends on them. Most housing-data mistakes are not arithmetic mistakes. Someone calculates a percentage correctly, but uses the wrong denominator. Someone compares two values accurately, but one describes a five-year period and the other a month. Someone finds the right property, then mistakes the owner’s mailing address for the property’s location. A polished chart does not repair those errors. This handbook is a practical way to keep the question, evidence, and decision together. Its examples focus on Hillsborough, Pinellas, Pasco, and Hernando counties. That is a defined study area, not a claim that these counties are the only places people mean by Tampa Bay. Readers working in Manatee, Sarasota, or another nearby county can use the same methods after replacing the local offices and geography codes. Do not silently add those counties to a table labeled with the four-county study area. A number needs a passport Before using a number, give it a short record of its identity. Include the source agency, dataset, table or field, geography, reference period, unit, universe, and retrieval date. When available, include the uncertainty measure. The universe means the population being counted: housing units, occupied housing units, renter-occupied units paying cash rent, or something else. Different universes cannot be substituted simply because the labels sound close. For example, the Census Bureau’s QuickFacts page for Hillsborough County reports median gross rent of $1,667 for 2020-2024. That observation has a geography, a period, a dollar unit, a housing-cost concept, and a source. It is not a claim that the typical apartment advertised in October 2026 costs $1,667. The first statement is supported by the published table. The second requires different evidence. Citation C one. A useful citation therefore reads more like a label on a storage box than a slogan: “Census QuickFacts, Hillsborough County, median gross rent, 2020-2024, retrieved October 22, 2026.” A reader should not need to guess which version you used. Keep the source URL with the record. If the page later updates, your archived input explains why your old report differs. This takes less time than reconstructing a missing source after an owner asks where a figure came from. A short source register is also a useful filter. If you cannot fill in the universe or time period, you may not yet understand the number well enough to use it. Four questions that look similar but are not “What do renters pay?” is a household-cost question. “What are landlords asking?” is a current listing question. “What can households afford?” combines costs and income. “What should this particular unit rent for?” is a property-specific decision requiring current comparable units and judgment. All four questions concern rent. They do not have the same answer. Public survey data are strong at describing broad conditions. They put a property decision into context. A household paying a high share of income toward rent may be under financial pressure even where the dollar rent looks low beside another county. An area with expensive owner-occupied homes may still include rentals at many different price levels. Neither fact determines a lawful or sensible action toward a specific resident. A parcel record answers a different family of questions. It helps identify the assessment record, recorded characteristics, exemptions, and tax-year values. A recorded deed helps investigate a transfer. Neither should be treated as a complete substitute for a title examination or survey. Administrative records exist for a purpose. Understanding that purpose is part of using them well. The practical sequence is to classify the question before selecting the source. A monthly leasing decision begins with the actual unit and recent, relevant comparisons. A county housing brief begins with a consistent county dataset. A tax forecast begins with assessment rules and the proposed ownership situation. A filing task begins with the office that receives the filing. This prevents a good source from being asked to do the wrong job. The difference between an indicator and an instruction An indicator describes a condition. An instruction tells someone what to do. Treating the first as the second is a common shortcut. “County rent is high” does not mean “raise this tenant’s rent.” “Vacancy is high” does not mean “drop the asking price.” “Values rose” does not mean “buy now.” The missing steps include the particular property, time frame, legal obligations, financial position, and quality of the underlying estimate. The book therefore uses public data as context, evidence, and a way to ask better questions. It does not offer investment recommendations or individualized legal, tax, insurance, or appraisal advice. A method can be repeatable without turning every decision into a formula. Consider a fictional manager assessing a renewal. The manager has a signed lease, current charges, an inspection record, and three recent comparable listings. County gross rent supplies background, but it does not replace those documents. The manager records what each comparison includes: utilities, bedrooms, condition, lease length, and concessions. If the proposed change is material, the manager separately checks current notice requirements and the actual lease. This is a decision file, not merely a screenshot of a median. The same discipline works for a buyer. An appraiser’s prior-year tax figure is a record of one assessment situation. A buyer needs to consider what might change after transfer and whether a prior owner’s exemptions apply. The old bill belongs in the file, but a forecast should be labeled a forecast and tied to the appropriate current local guidance. Define the place before comparing it “Tampa” can refer to a city, a postal address, a county context, or a loosely described market. Those are not interchangeable. A property with a Tampa mailing address is not automatically proof of location within the city’s corporate limits. A ZIP Code is a postal arrangement. A Census ZIP Code Tabulation Area, or ZCTA, is a statistical area. Chapter 5 addresses the difference in detail. Start every project with a written boundary statement. For a county comparison, list the counties. For a neighborhood brief, list the exact tract or other statistical boundaries used and explain why they approximate the neighborhood. If the decision concerns a specific address, begin with the parcel and verify the jurisdiction separately. Never assume a familiar place name resolves the boundary. There is no perfect geography for every question. A smaller area may better resemble the subject property but have a less precise survey estimate. A larger area may provide a steadier estimate but blur important differences. Choosing a geography means balancing relevance and uncertainty, then explaining that balance rather than hiding it. Do not use the apparent precision of a map to avoid the choice. A sharp polygon boundary may enclose an imprecise estimate. A colorful ZCTA map is not a parcel valuation map. The map should show what was measured and where; it should not imply a degree of local knowledge that the data do not contain. Give the calendar the same attention as the map A report published in 2026 may describe 2020-2024 survey responses. A record retrieved today may describe an earlier tax roll. A property sale may have a closing date and a later recording date. Write both the reference date and the retrieval date when they differ. For multi-year survey estimates, the period is part of the measure. Calling a 2020-2024 estimate simply “2024 rent” encourages the reader to interpret it as a snapshot of December 2024. The estimate pools information over a period. Later chapters explain why adjacent five-year releases overlap and why comparisons need care. Citation M one. An easy filing rule is to keep three date fields: reference period, publication or update date when supplied, and date collected. Use an additional event date for transactions such as a deed. These fields answer different questions. “When did it happen?” is not the same as “When did we learn about it?” If a source supplies no update date, say so. Do not convert the date printed by your browser into the date of the data. A report can be timely because it explains its lag, not because every number appears recent. A fictional decision file Suppose a small owner wants to compare renewal conditions for two apartments. Both addresses are fictional, and no local market facts are implied by the example. Apartment A has a proposed monthly base rent of $1,600 with water included. Apartment B has a proposed monthly base rent of $1,550 with the resident paying water separately. Looking only at base rent makes B appear cheaper. If the question is household housing cost, the water treatment matters. The decision file should begin with the question: “Compare the total quoted monthly cost for a new twelve-month lease, excluding refundable deposits, and identify items requiring separate verification.” It then lists the known amounts, missing utilities, recurring mandatory fees, and any concession conditions. An advertised free month should not be treated as a permanent monthly reduction without stating the calculation and lease term. The file also distinguishes observations from assumptions. “Listing states water included” is an observation about the listing. “Water costs $50” is an assumption unless supported by the relevant evidence. An effective-rent calculation can be correct while its inputs remain uncertain. Labeling those inputs lets the next person improve the file rather than trust an unmarked guess. A county median may then be added as broad context, with the explicit warning that it is not a matched comparable. The result is more useful than either a bare listing comparison or a bare public-data statistic. Each source contributes what it can support, and none pretends to answer the whole question. Build the smallest useful workbook A practical housing workbook can start with five sheets: questions, sources, raw data, calculations, and findings. The questions sheet defines the decision and boundary. The sources sheet records where the evidence came from. The raw-data sheet preserves the input. The calculations sheet makes derived measures visible. The findings sheet states what the evidence supports and what remains unknown. Do not overwrite imported values to make a chart look right. If a column needs cleaning, create a cleaned column or a separate transformation. Retain the original label beside the new label when a recode could change meaning. Blank, zero, suppressed, unavailable, and not applicable are different states. A workbook that converts all of them to zero can produce false comparisons without any visible spreadsheet error. Add a short checks sheet when the project grows. Verify row counts, duplicate keys, expected county names, numeric ranges, and consistency of periods. A check that fails should produce a visible warning, not disappear in a hidden column. The point is not to build elaborate software. It is to catch a mistake before it reaches a reader or a decision. A written finding should be short enough that someone can challenge it. “County A has a higher published median gross rent than County B in the same 2020-2024 QuickFacts release” is testable. “County A is a hotter market” imports a broad interpretation that requires more evidence and a definition of “hotter.” Plain words make analytical claims easier to audit. Keep public analysis separate from private records A market workbook and a resident ledger have different privacy needs. The fact that a property address can be found publicly does not make a resident’s payment history suitable for publication. Use separate files, permissions, and export rules. A public example should use fictional names and invented transaction details explicitly identified as fictional. When a project combines a public parcel file with private operational records, treat the combined file according to the private material it contains. Removing a name may not be enough if an address, unit number, and unusual event identify the person. Before sharing, ask what the recipient actually needs. A county count can often answer the question without a row-level export. The practical gain is also organizational. People doing daily administration should not have to wonder whether a spreadsheet is safe to email. A clear label such as “public analysis, no resident records” or “restricted resident ledger” prevents accidental reuse. Privacy is easier when it is built into the file structure rather than added at the end. What a good result looks like A good result is not the longest report or the most colorful chart. It is a finding someone else can reproduce, with enough context to avoid using it incorrectly. It states the question, explains the source, preserves the calculation, shows the limits, and identifies the next appropriate check. Sometimes the best result is a narrower answer. You may be able to compare published county medians but not rank five neighborhoods reliably. You may identify a parcel but still need a title professional to settle an ownership issue. You may establish that a flood map is effective for an address while still needing an insurer’s current quote and the relevant policy terms. Those limits are useful findings, not failures to finish. The following chapters build from this foundation. Read the source and uncertainty chapter before relying on a comparison. Read the recordkeeping chapter before designing a private ledger. When you return months later, begin with your source register rather than the old chart. The old chart shows what you concluded. The register explains whether that conclusion can still be used. Chapter practice Choose one housing number you have recently seen. Write its full identity without looking at the chart title alone. Can you identify the universe, geography, period, unit, and source? If not, recover them before interpreting it. Then write one supported statement and one tempting but unsupported statement. The difference between those sentences is the central skill this book teaches. Sources and verification Citation C one. U.S. Census Bureau, Hillsborough County QuickFacts, median gross rent, 2020-2024. Retrieved October 22, 2026. H T T P S colon slash slash www dot census dot gov slash quickfacts slash fact slash table slash hillsboroughcountyflorida slash HSG860224 Citation M one. U.S. Census Bureau, American Community Survey 5-Year Data. Geography availability and comparison profile description. H T T P S colon slash slash www dot census dot gov slash data slash developers slash data hyphen sets slash acs hyphen 5year dot html