Economic research, built on a platform that already holds the data
We publish standing reports on housing, labour and regional economies, and we take on custom work for clients who need an answer the published research does not contain. Every figure comes from federal data this site also gives away for free, so any conclusion we reach can be checked against its own source.
The practice exists because the hard part of applied economics is rarely the analysis. It is the assembly: acquiring the data, reconciling the geographies, understanding what each series does not say. That work is already done here, standing and versioned, which is why our engagements start where a conventional study finishes.
Published research
Standing reports, each sold once and owned permanently. The opening sections and the complete methodology of every one are free to read before you decide.
Where tax-credit housing pencils: maximum LIHTC rents against market rents, income-qualified renter demand and the supply gap, for 1,551 counties on FY2026 HUD income limits.
For: Acquisitions and development teams at LIHTC, workforce and build-to-rent housing developers, their syndicators and lenders, and the housing agencies and consultants who screen the same markets.
- Published
- August 7, 2026
- Sections
- 18 (15 paid)
- 275 pages
- Read time
- 65 min
Where to build next: permits, population, prices and school demand for every U.S. county of 50,000+ and every permitting city, ranked for residential developers.
For: Land acquisition and market-selection teams at homebuilders, multifamily and build-to-rent developers, and the capital that funds them.
- Published
- August 7, 2026
- Sections
- 18 (15 paid)
- 180 pages
- Read time
- 60 min
Where America's people, jobs, homes and paychecks actually moved between 2020 and 2026: every state ranked, every claim sourced, down to the neighborhood.
For: Site selectors, economic developers, relocation and expansion advisors, market analysts, and policy staff who need one defensible national reference.
- Published
- August 6, 2026
- Sections
- 22 (19 paid)
- 159 pages
- Read time
- 55 min
Reports are not on sale yet.Checkout hasn’t launched, so nothing here can be bought today. The free sections of every report, including the full methodology, are readable now, and each report’s page will take your address if you want to be told when it opens.
Who does the work
Bishoy Mikhail
Economist
Bishoy is an economist and the author of the research published here. He built pumsdata to close a gap he kept running into in applied work: the federal statistical system publishes an enormous amount of high-quality data, and almost none of it is usable without first spending weeks turning it into something you can ask a question of.
His work runs from weighted microdata analysis and index construction to housing market and regional economic studies. The engineering behind this site is his as well, which is unusual for an economist and is the reason the research can be priced the way it is: the analysis and the data infrastructure are not two vendors billing separately.
He is opinionated about method in one direction only. Every report he publishes carries a methodology section outside the paywall, states the universes and floors it used, and names the things its sources cannot support. When the data will not answer a question, the answer is that it will not.
Write directly: research@pumsdata.com
What we take on
Engagements run from a two-day screening question to a multi-month study. The common thread is that the data work is already standing, so the time you buy is spent on your question rather than on assembly.
Market screening and site selection
Which markets survive a first cut, and why. County- and city-grain screens across permits, rents, income limits, migration and supply, run on the same engine behind the published reports, scoped to your pipeline rather than ours.
Applied econometrics
Weighted microdata cross-tabulation, regression and index construction on ACS PUMS and the federal series: elasticity and demand estimation, shift-share and decomposition analysis, composite index design with published weights, and small-area estimation where the sample allows it.
Economic and fiscal analysis
Regional economic profiles, labour-market and wage analysis, industry concentration and exposure studies, housing supply and affordability assessments, and the impact work that goes in front of a board or a public body.
Survey design and polling
Instrument design, sampling frames built from the microdata rather than guessed at, weighting and post-stratification to known population margins, and analysis that reports its own margin of error rather than burying it.
Custom datasets and pipelines
When the answer needs a dataset nobody publishes: acquisition, cleaning, geographic reconciliation and a documented pipeline you keep. Every dataset behind this site was built this way.
Expert review and second opinions
Reading someone else's study, model or forecast and saying plainly what it supports and what it does not. Often the cheapest engagement we do, and occasionally the most valuable.
Where the name comes from
PUMS is the Census Bureau’s Public Use Microdata Sample: the anonymised, person-by-person and household-by-household records behind the American Community Survey. It is the rawest form in which the United States publishes data about itself. Rather than a table someone else decided to build, it is several million individual records with survey weights attached, from which any table can be built.
It is also, in practice, the least accessible thing the statistical system produces. Using PUMS correctly means downloading raw files, knowing that a record represents hundreds of people through a replicate weight, understanding which universe each variable applies to, and owning statistical software to do any of it. The data has always been free. Reading it has not.
pumsdata is named after that dataset because opening it is the thing the platform did first: weighted cross-tabulation of federal microdata, correct by construction, without writing code. What began there now spans dozens of federal families — permits, rents, income limits, employment, prices, schools, migration, output — assembled so they can be asked one question together.
We are careful about the claim we make here. The data is public and we hold no exclusive access to any of it; anyone determined enough can rebuild what we have. What we did was remove the practical barrier — and then keep doing it, across enough sources, for long enough, that asking a real question of them costs minutes instead of months. That is the whole advantage, and it is the reason the research is priced the way it is.
How we differ
Why is your work priced below what a research firm charges?+
Because most of the cost of a study is assembly, and ours is already done. A conventional engagement spends its first weeks acquiring data, reconciling geographies, and rebuilding the same national baselines from scratch — billed hourly, every time, for every client. pumsdata already holds those datasets, cleaned, joined and gated: dozens of federal families at state, metro, county, place and neighbourhood grain, refreshed on their own release calendars. We start where the usual engagement finishes. What you pay for is the judgement, the customisation and the writing, not the plumbing.
Doesn't a lower price mean less rigour?+
The opposite, in the part that matters: because the pipeline is standing and versioned, the same figures can be re-derived on demand and checked against the free public pages on this site. Every published report states its universes, floors, exclusions and known limitations in a methodology section that sits outside the paywall, so a reader can attack the method before paying for the conclusions. Most studies at ten times the price do not publish that.
What makes the data itself different?+
Not exclusivity — nearly everything here is public. The difference is that public and usable are not the same thing. The Census microdata behind this platform ships as raw records with survey weights and requires statistical software and the knowledge to weight it correctly; the affordability tabulations arrive as hundred-megabyte files with columns named T8_EST69. The work was making that usable and keeping it correct, then doing it again across dozens of sources so they can be asked one question together. That assembly is the asset.
Do your reports replace a formal market study?+
No, and we say so inside them. The published reports are market-selection instruments: they tell you which markets are worth the cost of a full study and which are not. A tax-credit application still needs a commissioned study from a qualified analyst. Screening twelve markets to find the three worth studying is where we save you money.
Can we license a report for a whole team, or commission a custom edition?+
Yes to both. A single report purchase is licensed to the buyer and their immediate team; organisation-wide licences and white-labelled regional editions — your metro, your branding, the same pipeline — are quoted individually. The marginal cost of a new geography is low, which is reflected in the price.
What will you not do?+
We do not produce a number to support a conclusion someone has already reached, and we will tell you when the data cannot answer the question you asked. Several sections of the published reports exist specifically to say what the sources do not support. That is the product.
Work with us
Most engagements start with a short conversation about what you are trying to decide. If the published research already answers it, we will point you at the section and charge you nothing.