Security guard wages by metro area: every percentile, delivered

Two different products carry that label, and mixing them up is how guard-rate decks go wrong. Occupational wage distributions estimate what the job pays: one row per area x year-quarter x industry x occupation, carrying an employment count plus a mean, a median and the 10th-through-90th percentile rungs, doubled into hourly and annual variants. Employer payroll aggregates measure what the industry spends: one row per quarter x area x ownership class x industry code, carrying establishment counts, employment and total wages tabulated near-census from payroll records rather than estimated from a sample. The first prices the guard; the second tracks the guard-services firm.

Datadory delivers both as typed rows in one joinable extract: Occupational Employment and Wage Statistics (OEWS) - support occupations as the distributional core, Quarterly Census of Employment and Wages (QCEW) - NAICS 56 as the payroll instrument, and the structural and cross-border records that say who sets those rates and how concentrated the local vendor bench is. Every figure ships with its geography, occupation code and reference period attached, so nothing arrives as a bare number.

The occupational core resolves every estimate along three axes at once - where, which industry, which occupation - then attaches the money. Geography runs from California statewide totals down through every metropolitan statistical area and the OES survey regions used for sub-metro tabulation. Industry cuts separate 'Total, All Industries' from named NAICS lines, which is what isolates in-house guard departments from contract-security vendors. Occupation follows the federal SOC taxonomy, and one code does nearly all the guard work: 33-9032 Security Guards, with 37-2011 Janitors and Cleaners and 49-9071 Maintenance and Repair Workers riding the same table whenever a bundled facilities contract needs them.

The payload is a ladder, not a single number: a mean, a median and the 10th, 25th, 50th, 75th and 90th percentile rungs in whichever wage type the row declares - Hourly Wage or Annual Wage or Salary. Employment counts sit beside the wages, and precision travels with the estimate: a Mean Relative Standard Error column qualifies every mean, so thin cells in small metros announce themselves instead of masquerading as solid figures. Percentile cells are withheld where the sample underneath is too small to present honestly; absence means suppressed, never zero.

One row per area x year-quarter x industry x occupation cell, identical anatomy whether the geography is the state total or one metro. Four rows exactly as they arrive from Datadory, captured during the August 2026 review:

AREA_NAME   YEAR  QUARTER  INDUSTRY             SOC     OCC_TITLE                     WAGE_TYPE  EMP     MEAN_WAGE  MEDIAN_WAGE
California  2026  1st Qtr  Total, All Industry  339032  Security Guards               Annual     199480  46057.82   44335.26
California  2026  1st Qtr  Total, All Industry  372011  Janitors and Cleaners ...     Annual     234240  45329.83   42130.8
Chico MSA   2026  1st Qtr  Total, All Industry  372011  Janitors and Cleaners ...     Hourly     900     21         19.3
Chico MSA   2026  1st Qtr  Total, All Industry  371011  First-Line Supvrs, Housekpng  Annual     100     58351.52   61171.18

Reading them: statewide, just under 200,000 security guards average 46,057.82 dollars a year against a 44,335.26 median - a narrow gap saying the top-line distribution is tight, though the percentile columns underneath tell a wider story. Drop to a single metro and cells thin out: the Chico cut covers roughly 900 people in a sibling support occupation at 21 dollars an hour against a 19.30 median, exactly the scale at which the relative-standard-error column starts doing real work. The fourth row is the supervision premium in one line - 100 first-line housekeeping supervisors averaging 58,351.52 dollars while their median runs higher still at 61,171.18 - the same premium a guard rate card meets in its own 75th and 90th rungs.

Occupational titles longer than the column above ship untrimmed; they are shortened here only to hold the table's shape.

Because guard pay skews. Armed posts, clearance-cleared posts and supervisory contracts sit far above the bulk of routine unmanned positions, pulling the mean upward while the median holds its ground. When two metros' means look close, their medians and 75th-percentile rungs often are not - and the spread between the 10th and 90th rung inside one metro can be wider than the entire gap between metros.

For anyone setting or evaluating rates, the working rule: benchmark routine unarmed posts against the 25th-to-50th percentile band, and treat the 75th and 90th rungs as the territory of supervisory, armed or premium-clearance contracts rather than evidence that a metro is generally expensive. Quote the rung, name the wage type, stamp the reference period - a range survives negotiation; a lone mean invites it.

One caution worth repeating because it decides which instrument answers which question: these occupational estimates are snapshots, not a time series. The survey rotates establishments, so lining up consecutive years mixes changing samples. For movement over time, the payroll record below is the right tool.

The Quarterly Census of Employment and Wages answers a different question about the same workers: what employers actually pay in aggregate, tabulated near-census from unemployment-insurance payroll records rather than estimated from a sample. Its rows carry establishment counts, three monthly employment readings, total quarterly wages and average weekly wages for every county, ownership class and NAICS level from the all-industry total down to six-digit industries - Level 6 reaches 561612 Security Guards and Patrol Services, the exact frame a guard-rate question wants.

Because every worker covered by unemployment insurance appears in the tabulation, small metros that an occupational survey estimates thinly stay fully visible here. Cells that would identify a cooperating employer are withheld - read the gaps as disclosure decisions, not missing data.

Time works differently too: quarterly observations run 2004 through fourth-quarter 2025, twenty-two years deep, in recent-period files of roughly 815,000 rows - long enough to compute year-over-year moves in guard-firm payroll, which the occupational snapshots explicitly decline to support. Use the percentile ladder to price the occupation; use this record to track the industry's payroll behind it.

A wage number lands differently depending on how fragmented the local vendor bench is, and Statistics of U.S. Businesses (SUSB) measures exactly that. It counts firms, establishments, employment, annual payroll and receipts by six-digit NAICS and enterprise employment-size class down to county and congressional-district geography, annually from 1997.

Its own 2022 United States file explains why fragmentation matters in this trade: NAICS 5617 Services to Buildings and Dwellings - the classification bucket housing much of the guard-and-facilities vendor world - records 215,428 firms with about 2.1 million employees, and 149,120 of those firms employ fewer than five people. A metro dominated by that micro-firm class quotes, bids and holds rates differently from one where national chains hold the establishment count. One reading quirk worth knowing: receipt figures populate only in years ending in 2 and 7, so a receipts series steps every fifth year while payroll and employment run annually.

Pairing rule: take the percentile ladder for the rate card, the payroll record for direction of travel, and size classes for who is actually across the table.

Yes, at industry and company grain rather than occupation grain - label the difference.

Statistics Canada - NAICS 56 data tables tracks monthly payroll employment for the support-services sector across all fourteen provinces and territories: 17,598 series and about 5.3 million datapoints running January 2001 through May 2026, with explicit members for 5616 Investigation and security services and 5617 Services to buildings and dwellings. Monthly, quarter-century deep and province-cut, it is the instrument for asking whether Canadian guard-firm employment is expanding ahead of a bid.

The UK Companies House register answers the supplier-density half: 44,686 companies carry SIC 80100 private security, alongside 20,302 in combined facilities support and 63,603 in general building cleaning, live and dissolved together so survival and churn stay computable. Two cautions travel with it - SIC codes are self-declared by directors rather than verified, and the multi-code nature-of-business array matters, since a firm declaring both facilities and cleaning codes belongs in both markets.

The honest limitation: neither record publishes an occupation-level guard wage by city the way the American occupational table does. Cross-border work pairs payroll series and company counts with occupational detail only where a national statistical office presents it - so stamp every cross-country figure with its actual grain.

Five records, five different questions - and choosing among them decides what a guard-rate claim can assert. The table below lays out the layers; the deeper pairing argument between an occupational wage ladder and a supplier register is argued line by line in OEWS support occupations vs UK Companies House registrations.

Two failure modes account for most bad guard benchmarks. Quoting the mean where the 25th-percentile rung answers the question - armed-post averages smuggled into an unarmed rate card. And reading average weekly wages from the payroll record as if they were occupational rates, when they are industry-wide payroll averages covering every worker in the cell, janitors and dispatchers included. Both mistakes disappear the moment each figure carries its layer label.

Six kinds of teams get outsized value, and each works a different corner of the stack.

Procurement and facilities buyers set and defend posted hourly rates metro by metro, quoting the 25th-to-50th band with the reference quarter named. Security-services operators benchmark bid rates in every territory they sell into and pair wage ladders with establishment counts to read local competition - the workflow map lives on diversified support services data for competitive intel product teams. Market researchers and consultants build city-rank tables clients republish, where relative-standard-error columns decide whether the deck survives scrutiny - see diversified support services data for market researchers. Investors and quant researchers read guard-wage drift as a forward input on labor-intensive services margins - see diversified support services data for investors and quants. Data scientists and analysts anchor compensation and unit-economics models on the percentile ladder rather than a lone mean, joining across years on SOC code. Journalists, academics and students cite figures whose method, vintage and error structure publish alongside the numbers, which makes every citation defensible in print.

Because the hard part was never learning these records exist - it is keeping five heterogeneous series joinable. Occupation codes resolve onto one taxonomy, geography names arrive keyed and consistent across the occupational and payroll layers, industry cuts stay intact instead of collapsing into blends, percentiles ship as seven columns rather than one summary statistic, and precision plus reference period ride beside every value they qualify.

Delivery is yours to set: API, files, or straight into your warehouse - daily, weekly, or hourly. Tighten the interval while a contract cycle is live, loosen it once the study turns historical; switching later is a settings conversation, not a migration. Start with a sample: name the metros, the occupation codes and the wage types, and rows shaped exactly like the field dictionary arrive sized to test in your own pipeline.

Guard wages by metro are one question inside a wide support-services pool. Start with the diversified support services data hub, which ranks every record in the slice and shows where the wage question sits among them. Then go to the products themselves: Occupational Employment and Wage Statistics (OEWS) - support occupations for the percentile ladder and Quarterly Census of Employment and Wages (QCEW) - NAICS 56 for the payroll instrument. OSHA Enforcement & Injury Data adds the employers' safety record, the diversified support services data guide maps the wider pool with quality scores and use cases, and the best diversified support services datasets ranking settles where to start. Vocabulary lives on the OEWS and QCEW glossary entries.