The Dallas Fed's "SF Wages" Quote Is a Sourcing Map, Not a Warning

The Sept 3, 2026 Beige Book buried a line every founder hiring in Texas should read twice. A Dallas district contact told the Fed that job candidates are asking for "San Francisco wages" even as Texas wage growth is decelerating and manufacturing workers are migrating in from California. Read as a complaint, it sounds like entitlement. Read as a sourcing signal, it is the clearest map of arbitrage the Fed has printed in two years.

What the Beige Book actually said

The Dallas district reported a skills, geography, and wage-expectations mismatch, and one contact specifically cited candidates anchoring pay to Bay Area bands. That is the news; the rest is context.

The exact language from the Sept 3, 2026 release, drawn from the Dallas Fed's Texas Business Outlook Survey:

• "Numerous instances of labor market mismatch in terms of skills, geography, and expected wages."
• "One contact noted job candidates asking for 'San Francisco wages,' while another stated migration of skilled manufacturing workers from California helped ease regional labor shortages."
• Texas execs' top three hiring blockers: lack of available applicants, applicants looking for more pay than offered, and lack of technical skills.

The national frame in the same Beige Book roundup: "Wage growth was modest to moderate in most Districts. Significant wage increases were most often connected to demand for skilled workers in construction and manufacturing." Translation: tech is no longer the wage-growth driver. The SF anchor Dallas candidates are quoting is a story about reference points, not about a hot local market.

Why the "SF wages" ask is rational, not entitled

Location-agnostic pay is now a named model, and it leaks into every candidate's reservation wage. That is the mechanism. Blame the pay policy, not the applicant.

GitLab, Basecamp, Cloudflare, and Automattic have paid the same salary regardless of geography for over a decade. Stripe, Shopify, Airbnb, and Discord run hybrid or partially remote bands that transmit similar anchors across metros. When a $170K senior engineer offer is $170K in San Francisco or rural Montana, every Dallas candidate who reads a public comp thread now walks into a local negotiation with a portable number in their head.

That number does not care about the Dallas Fed's regional wage survey, which pegged 2025 Texas wage growth at 3.5% (down from 4.3% in 2024) and projected 3.3% for 2026. The candidate's anchor was set by a public salary band at a remote-first employer, not by the local labor market.

The SF wage ask is a symptom of location-agnostic pay bands leaking into every candidate's reservation wage.

The arbitrage is bigger than the anchor

A Dallas engineer paid 90% of an SF band ends up with roughly 1.7x the real purchasing power of their SF peer. That is the number founders keep failing to run.

The math, using published benchmarks:

So the SF ask in Dallas is roughly a 26% premium over what the local employer market clears at. But because Dallas is 43% cheaper, an engineer accepting even 80 to 90% of an SF band in Dallas is dramatically overshooting parity on real terms. Founders paying full SF wages to a Dallas hire are effectively over-compensating by ~40% on a purchasing-power basis, unless that candidate has a live SF offer in hand.

That last clause is the whole game. The right counter to "SF wages" is not "no." It is "show me the competing offer," and then a band calibrated to the candidate's actual alternatives.

1.76x

Where the engineers actually live

Nine of the top ten metros for senior Python engineers in the US are not San Francisco. That is the single fact that reframes the Dallas Fed anecdote.

In Refolk's index of professional profiles, there are approximately 46,083 US-based Senior and Staff Software Engineers who list Python as a skill. The top hiring-market regions in that pool skew across San Francisco, New York, Austin, Boston, Chicago, and Lehi (UT), with meaningful clusters in Boca Raton, Fort Myers, Tempe, Minneapolis, Raleigh, and Brewster (NY). SF is one region among ten, not the market.

For AI and ML engineers, the story is tighter but still not SF-only. Refolk's index shows roughly 14,776 US-based AI/ML engineers, with top current-employer regions in San Francisco, New York, Brooklyn, Minneapolis, Boston, Rochester (NY), Tempe (AZ), and Greater Seattle. Even the scarcest talent class has non-SF depth that public salary guides never surface.

46,083

The reason smaller metros rarely appear in public comp aggregators is sample size: they only report where they have enough submissions to publish a band. Lehi, Boca Raton, Rochester, and Tempe fall below those thresholds, so the candidates in those metros do not see a local anchor and often default to the SF number by exposure. That is a sourcing opportunity, not a pay problem.

If you have been running the same "Bay Area senior backend, open to remote" search for a year, you are competing on exactly the anchor Dallas candidates just cited to the Fed. Describing the geography and the current-employer profile in plain English is the exact gap Refolk closes: you ask for the engineer you actually want, not the one a keyword filter set can approximate.

The mismatch is skills and geography, not pay

The Dallas Fed's own contacts flagged lack of available applicants and lack of technical skills alongside pay expectations. The fix is sourcing farther, not paying more.

Read the Texas Business Outlook Survey answer set in order: applicants unavailable, applicants asking for more, applicants lacking technical skills. Two of the three are supply problems disguised as pay problems. If the local pool is thin, the candidates who do surface get to name their price, and the ones they name come from wherever comp is loudest, which right now is San Francisco.

The response most founders reach for (raise the band) reinforces the anchor. The response that works:

1. Expand the geographic aperture past the three metros your ATS already knows.
2. Filter for engineers whose current employer is not a remote-first, location-agnostic payer.
3. Interview against a comp band calibrated to the candidate's live alternatives.
4. Reserve the AI/ML premium (a documented 15 to 25% wage bump per 2026 comp guides) for roles that genuinely need it.

That last point matters. Mid-career US software engineers (3 to 8 YOE) are seeing 2026 total compensation between roughly $160K and $390K depending on level, location, and equity. The AI premium is real. But paying it to a generalist backend hire because the candidate said "SF wages" is a category error.

The one category where the SF ask is defensible

AI and ML engineering. Everywhere else, the anchor is negotiable.

With only ~14,776 US AI/ML engineers in Refolk's index and a documented 15 to 25% wage premium on top of already-high SWE bands, this is the one talent class where "SF wages" reflects actual scarcity rather than remote-era anchoring. If a candidate can point to a live offer at an SF band from a frontier AI employer, that is not anchoring, that is the market.

For generalist backend, data engineering, mobile, and infra roles, the same ask is negotiable. The tell: ask the candidate what their current base is and what a competing offer looks like. If the answer is a comp thread rather than a written offer, you are negotiating against a reference point, not a market.

What to change in your sourcing this quarter

Stop searching by title and metro. Start searching by employer graph, current-comp anchor, and skill depth. That is a description problem, not a filter problem.

Practical shifts that move the needle on secondary-metro sourcing:

• Name the metros explicitly. Lehi, Tempe, Minneapolis, Raleigh, Rochester, Boca Raton, Fort Myers. These rarely appear in public city rollups but do appear in Refolk's index as real employer concentrations.
• Filter by current-employer pay policy. An engineer currently at a location-agnostic remote employer has already been paid the SF anchor. An engineer at Databricks, GitHub, Adobe, Notion, or Ramp in a secondary metro has not necessarily been.
• Reverse the pitch. Instead of leading with comp, lead with the specific work. Secondary-metro engineers, in aggregate, take fewer interviews per quarter and screen harder on team and problem than on band.
• Use the AI premium sparingly. Reserve it for roles where the candidate's LLM, RAG, or eval experience is load-bearing. Do not spend it on Python-plus-a-notebook.

The workflow this implies is a plain-English one: "senior backend engineers with 5+ years of Python and Postgres, currently at a product company in DFW, Austin, Lehi, or Tempe, not at a remote-first employer." That is the kind of ask that surfaces the engineers whose comp anchors are still set locally.

The bigger frame

The Beige Book quote is a leading indicator that comp anchoring has decoupled from local labor markets. That is bad news for founders who benchmark to metros and good news for founders who benchmark to candidates.

Sourcing in 2026 is a purchasing-power arbitrage as much as a talent search, and the founders who run the math beat the ones who match the anchor.

FAQ

Is the "SF wages" ask actually widespread or a one-off Dallas Fed anecdote?

It shows up as a single contact quote in the Sept 3, 2026 Beige Book Dallas district writeup, but the mechanism (location-agnostic pay bands at GitLab, Basecamp, Cloudflare, Automattic, and hybrid Stripe/Shopify/Airbnb/Discord teams) is national. Expect the same anchor in Austin, Raleigh, Minneapolis, and Tempe negotiations. The Fed just gave you the first named citation.

How do I tell if a candidate's "SF wage" ask reflects a real alternative or just an anchor?

Ask for a written competing offer, not a comp screenshot. If they have a live SF offer from a remote-first or hybrid employer, the ask is a market signal and you should either match, differentiate on non-comp, or pass. If they cannot produce one, you are negotiating against a reference point, and a band calibrated to their current base plus 10 to 20% usually clears.

Where should I actually source engineers to avoid the SF anchor entirely?

Metros with real senior engineering density that fall below public comp-aggregator thresholds: Lehi (UT), Tempe (AZ), Minneapolis, Raleigh, Rochester (NY), Boca Raton, and Fort Myers, alongside the obvious Austin and DFW. Refolk's index shows meaningful senior Python and AI/ML populations in all of these, and candidates in those metros typically have comp anchors set by local product companies rather than by remote-first public bands.

Does this arbitrage still work for AI and ML hires?

Less so. With ~14,776 US AI/ML engineers in Refolk's index and a documented 15 to 25% AI wage premium on top of already-elevated SWE comp, the SF ask for genuine AI/ML roles usually reflects actual scarcity, not anchoring. Reserve the premium for candidates whose LLM, RAG, or eval experience is load-bearing to the role, and negotiate normally for everything else.