How we work · Our approach
An evaluation of a reform era, not a second ratio book.
The City measured its baseline in 2021 and adopted its reform package the same year. Five years on, the update's job is to find what moved, what did not, and why, and to re-found the goals program on current evidence built to the 2026 legal standard.
The disparity index: a group's share of contract dollars (utilization) divided by its share of firms in the market (availability), with values under 0.80 marking substantial underutilization. Frederick already has a rigorous version of this number. The City's 2021 study, presented in March 2021, analyzed $91,192,290 in prime spending for FY2014 through FY2018 across a Washington-Baltimore-Arlington market area drawn to capture at least 75% of prime award dollars. It found statistically significant MWBE underutilization in all five procurement categories: Construction at an index of 0.70 (availability 41.37% against utilization 29.01%), Architecture & Engineering at 0.39, Other Services at 0.24, and Professional Services and Goods at 0.00, with zero prime MWBE dollars across all five years in those last two categories. Headline MWBE prime utilization was 25.09%, but 96.58% of those dollars went to a single Asian American-owned construction firm. HSG has extended the surface read with current public availability data: in the ABS 2022, minority-owned firms are 15.35% of Frederick County employer firms and women-owned firms 20.29%, both up from 2017, and the county's 25,500 nonemployer firms are far more diverse still (35.29% minority-owned, 43.14% women-owned). What no public source provides is a current utilization series: the only published program figure since the study is the City's first annual review, which reported roughly 20% of 2022 procurement spend ($8.8M of $44M) going to small, women-, and minority-owned firms combined, a different measure than the study's category grid.
It shows the size and direction of the gap between presence and participation, and it shows where to look first. Frederick's 2021 baseline was unusually decisive: significance in every category, two categories at literal zero prime MWBE dollars, and a concentration finding (one firm carrying 96.58% of MWBE prime dollars) that means the headline 25.09% overstated how broadly participation was shared. The City acted on it: the aspirational goals it adopted (Construction 41, A&E 40, Professional Services 32, Other Services 29, Goods 16) sit at the study's measured availability levels rounded, so the program's targets are the study's own numbers. The availability side has moved since: the county's minority employer-firm share grew from 10.67% in 2017 to 15.35% in 2022, and the women-owned share from 17.64% to 20.29%. A growing base against fixed goals is itself a finding in waiting. Together the baseline grid and the current availability signals are a credible, transparent first read of where Frederick stands.
Computing the ratio takes minutes, and the City has already paid for the ratio done properly once. An update that re-runs the same surface arithmetic on fresh data and stops there tells the City what it mostly already knows: a substantial, significant gap existed in FY2014 through FY2018, programs were adopted, and the base of diverse firms has grown. It would also miss the actual question. Frederick has run a Small Business Reserve, an MWBE goals program, and annual-review requirements for nearly five years. The update the City needs is an evaluation of that reform era: what moved, what did not, which mechanism produced which result, and whether the goals, still resting on FY2014-FY2018 data, are founded on evidence that would hold today. We treat the ratio as the first step and build the evaluation and the causal analysis behind it, so the City receives recommendations it can act on and defend.
Where we go
The ratio is the start, not the answer
Start with what the record already holds. The City's 2021 study found statistically significant MWBE underutilization in all five procurement categories for FY2014 through FY2018, with zero prime MWBE dollars in Professional Services and Goods and 96.58% of all MWBE prime dollars concentrated in one firm. Its anecdotal record described an entrenched informal network, repeat use of the same firms, unmonitored good-faith efforts, and weak outreach. The City responded within the year: resolutions in November and December 2021 amended the Purchasing Policies to create the Small Business Reserve and the MWBE program, set annual aspirational goals at the study's measured availability levels, and required yearly effectiveness reviews, with a Contract Compliance Administrator to run the programs and a DEI Advisory Board reviewing procurement annually.
A baseline and a reform package tell you where things stood and what was tried. They do not tell you what the reform produced, because the only public performance figure since is a single 2022 review (roughly 20% of spend to small, women-, and minority-owned firms combined, 154 newly certified small businesses, 60 exemptions), and no later review has been published. The deeper work treats the 2021 package as a dateable intervention and evaluates it mechanism by mechanism: did the SBR band move small-firm participation, did contract-specific construction goals move MWBE subcontracting, did certification growth translate into award growth, did the informal-network pattern the study described in FY2014-FY2018 survive into the reform era? It traces the procurement funnel stage by stage, controls for capability, and rebuilds availability to the current market. And it confronts the data architecture directly, because the 2021 study found zero tracked subcontractors in FY2018, and Frederick County's March 2026 study next door showed what that gap costs: substantial measured disparity that could not support a program because only 27% of prime contractors had kept subcontractor records. The update's value depends on fixing that architecture, not just recomputing ratios across it.
Pinpointing the cause is what lets a recommendation be specific. Once the analysis shows which stage of the funnel the drop-off happens at, and which explanation holds up, the fix follows from it. A drop-off at registration or certification points to outreach and a simpler path across the City's four recognized certification routes. A drop-off at awareness points to the City's monthly procurement forecast and how notices reach firms. A drop-off after award, in slow payment, points to payment practices. A data gap points to the OpenGov platform the City has already bought, which can carry award identifiers, clean commodity coding, subcontractor tracking, and a public utilization dashboard the City does not yet publish. Each fix attaches to a mechanism the City can change, and most of what closes these gaps is open to every firm and does not turn on the race or gender of its owner. That sequencing is not just prudence in Maryland: the State's own 2026 reauthorization statute codifies that race-neutral efforts will be used to the maximum extent feasible, with race-conscious measures only where necessary. The goal is more firms from every segment of the community competing for and winning City work.
The discipline that makes a finding credible
What a gap can and cannot tell you
A low disparity index does not prove the City did anything wrong, and we do not start from that assumption. A raw difference between availability and utilization is a correlation, and a correlation can have several sources, only some of which point back at the City. The cause may be something the City controls, such as how quote lists are built, how the SBR band is calibrated, or how bonding and experience thresholds are set. It may be marketplace-wide, reaching across public and private buyers and rooted in access to capital or in the informal networks that decide who gets the first referral. It may be a legitimate, non-discriminatory difference in the firms themselves, in size, capacity, or the lines of work they are positioned to bid. Our job is to test those explanations against the evidence. We control for capability and qualifications, we examine whether the City's own criteria are appropriate and least restrictive, and we build the record to the strong-basis-in-evidence standard the Supreme Court set in Croson. Frederick sits in the Fourth Circuit, where the controlling appellate case upheld race-conscious measures only for the groups whose evidence was statistically significant and struck them for the rest, so group-specific rigor is not a preference here, it is the governing law. The 2026 climate has only raised that bar: the courts increasingly demand evidence gathered before a remedy, not procured to defend one. A benign explanation is just as useful to surface, because it shows where effort is not needed. We report whichever answer the evidence supports.
The ladder of analysis
From a gap to its cause
Each step answers a specific causal question and addresses a rival explanation, so the study can say not only that a gap exists but why, and where the City can act. The tag on each step marks whether the cause it isolates is something the City controls, a marketplace-wide condition, or a legitimate difference in capability.
PhD-led stratified custom-census survey of firms in a spend-derived market area, weighted to the City's expenditure mix across its five procurement categories, telephone plus email, to a 95% confidence interval per industry subsector. Anchor willingness to active public-work registration and observed bid behavior. Because a firm must register to bid, the City's vendor rolls, the SBE certification roster, and the MDOT directory define a bounded, followable universe of firms that have signaled willingness to do public work, the tractable government-contracting market the survey then measures readiness within. Consistent with Croson's qualified-willing-and-able standard, do not capacity-discount the denominator, since current capacity is itself partly a product of past conditions; the only ability filter is genuine market-area presence and line-of-business match. Reconcile the new census against the 2021 study's availability estimates (the same numbers the City adopted as its goals) so change over time is measured on a consistent basis, and replace equal-weight ABS firm counts with this weighted denominator before recomputing any index.
Rules out the leading attack that the gap is an artifact of an inflated headcount denominator (counting firms that never bid or are out of line-of-business). If the gap persists against a willing-and-able, spend-weighted denominator, an availability-measurement explanation is ruled out. It also re-founds the goals program: goals set at FY2014-FY2018 availability get a current, defensible availability base.
Interrupted time-series analysis around the adoption date. Build a monthly or quarterly participation series by group and category from the City's transaction records, extend it back toward FY2019 where bridge records exist, and anchor the deeper baseline to the 2021 study's FY2014-FY2018 record. Test for level and trend breaks at the intervention date, use categories and dollar bands outside every program mechanism as comparison series to absorb market-wide trends, and attribute any break to its mechanism: the SBR band should move small-firm shares in the reserved band specifically, contract-specific goals should move construction subcontracting specifically, certification growth should show up first as certification and then as awards. Document the IonWave-to-OpenGov seam so a system migration is never misread as a participation change. This is the step that makes the update an evaluation rather than a recomputation.
Rules in or out program effect. A participation shift timed to the intervention, located in the mechanism's own lane, and absent from the comparison series is evidence the program moved the number; a flat series says the mechanism has not yet reached its target, which redirects the recommendation to why. It also rules out false credit: growth that predates the intervention or appears equally in untouched lanes is market trend, not program effect.
Build a bid-level dataset of who requested or downloaded documents, who submitted, and who won, by solicitation, joined across the IonWave and OpenGov platforms. Compute a bid-to-award ratio by ownership and compare it to the availability-to-award ratio. A gap that closes at the bid stage points to a supply or awareness problem upstream; a gap that opens at the bid stage points to a selection problem at evaluation. The 2021 baseline sharpens the question: zero prime MWBE dollars in Professional Services and Goods could mean no diverse firms bid in those categories or that bidders lost, and those are different problems with different fixes. Pair with lost-contract records where they exist so willingness rests on bid behavior rather than registration alone.
Separates the supply explanation (firms are not bidding) from the selection explanation (firms bid and are not selected). This decides which half of the funnel the rest of the analysis must concentrate on and rules one of the two out as the dominant channel, category by category.
Construct a stage-by-stage conversion funnel and compute the retention rate by ownership at each transition: registered share, certified share, opportunity-aware share, bidder share, responsive share, awardee share, paid-on-time share, and subcontract-participation share. Locate the stage with the largest ownership-specific drop. Each stage implicates a different cause and a different fix: a registration or awareness drop is a supply and outreach problem, a bid-to-responsive drop is a criteria problem (route to S6), a responsive-to-award drop is a selection or network problem (route to S7), and a payment drop is a cash-flow problem. Frederick's certification layer gets its own reading, because the City recognizes four routes (SBA certification, MDOT, USDOT, and City self-certification) and the funnel can show which routes firms actually use and where each one leaks.
Locates causality in process space. It rules out a single global story by forcing the gap to a stage. A supply-stage drop rules out a selection-bias claim; a payment-stage drop rules out an availability claim; a subcontract-stage drop points to prime behavior rather than City selection.
This is the engine at the center of the study. Decompose each group gap into an explained block (firm age, size, receipts, bonding capacity, prior experience, industry mix, bid frequency) and an unexplained residual. Estimate capacity-controlled but-for models of win probability conditional on bidding and of award dollars, correct for non-random selection into bidding with a two-stage model so the upstream decision to bid is measured rather than assumed away, and run within-firm comparisons across procurement methods so time-invariant firm quality is differenced out. Report every contested control (incumbency above all) both ways: with it as the conservative lower bound, without it as the read that treats incumbency as itself a channel of past exclusion. A residual at or near zero means capability differences explain the raw gap; a significant surviving residual is characterized as differential returns, never labeled discrimination by the statistics alone.
Rules in or out the legitimate-capability explanation. A raw index like the 2021 study's 0.39 in Architecture & Engineering can reflect that diverse firms are younger, smaller, or less capitalized on average. If the gap vanishes under controls, the cause is capability and the remedy is capability-building, not preference. A surviving residual, reported both ways on the contested controls, is the evidentiary core the Fourth Circuit's standard demands.
Inventory and code every City solicitation in the study period for each screen. For each criterion, estimate the share of the available minority-owned and women-owned pool versus the comparison pool that could satisfy it, using the census firm-age, capital, bonding-capacity, and experience data, and apply a disparate-impact screen by ownership using the four-fifths convention as an analytic analogy, flagged for significance. Run a fit-to-scope test (is an experience floor necessary for routine work; is bonding calibrated to real payment risk; is a contract bundled past small-firm reach). Audit the SBR's own boundaries, including the discrepancy between the band as stated on the City's 2026 program form and as described in its other published materials, and the exemption lane, where 60 purchases left competition in 2022 alone. Conclude with a least-restrictive-means review producing a keep, right-size, or replace recommendation per criterion. Augment the manual coding with an AI-assisted read of the full procurement-rules corpus that surfaces internal rule conflicts and candidate disparate-impact provisions for a named investigator and counsel to verify; the AI accelerates and broadens the review but makes no finding, and a responsible-AI advisor governs the bias-testing and documentation of that step.
Rules in or out the explanation that the City's neutral-looking requirements are themselves the barrier. A criterion that few minority-owned firms can meet, and that exceeds what the work requires, is a City-controllable cause distinct from any bidder behavior, and it grounds race-neutral remedies before any race-conscious measure is considered.
Measure repeat-player concentration (share of dollars to firms with prior awards and concentration indices by ownership), the share of dollars flowing through sole-source, small-purchase, exemption, and referral channels versus open competition, and prime-sub pairing patterns (do the same primes pair with the same subs, and are minority- and women-owned firms absent from those pairings). Map the co-bidding and prime-sub graphs, test whether ties form along ownership lines after controlling for neutral sorting factors (industry overlap, size, geography, bonding), and compare ownership-specific gaps in informal channels against open-competition channels within the same firm. Pair with the structured anecdotal protocol (S9) on network and referral themes, so the statistical signature and the experiential record are read together.
Rules in the social-capital channel as a cause and distinguishes it from formal-rule causes (S6) and from capability (S5). It supplies the theory a bare ratio lacks: if gaps concentrate in informal, relationship-driven channels and shrink in open competition, the durable barrier is network access, which both criteria reform and capability-building miss. It also directly tests whether the informal-network pattern the 2021 study documented persists in the reform-era data.
Estimate private-sector earnings and business-formation disparities for the regional market: linear regressions on wages and business-owner earnings and logistic regressions on self-employment and formation rates, using ACS PUMS microdata for the Washington-Baltimore region, controlling for education, age, capital proxies (homeownership, home value, mortgage status, unearned income), and market-area residence. Layer the ABS receipts evidence, which already shows the direction: average receipts for all Frederick County employer firms are $5.256M against $1.677M for minority-owned and $886K for women-owned firms. Add a dedicated access-to-credit sub-analysis from HMDA, Small Business Credit Survey, and survey-reported financing outcomes, and link capital constraints to firm size, bonding capacity, and the formation gap so the capital channel is explicit rather than folded into a generic capability control.
Distinguishes City-specific causation from marketplace-wide disadvantage. If the gap is large in the private economy too, part of the cause is a condition the City did not create, which under Croson can support the compelling interest while pointing remedies toward marketplace-facing tools like capital and bonding access. If the private-sector gap is small but the City gap is large, the cause is more local and City-controllable.
Run separately instrumented, structured interviews using distinct guides for businesses and for professional and trade organizations, recruited broadly (email, mail, personal contact, public sessions) to reduce self-selection. Code each account to the procurement-funnel stage (S4) and the barrier framework. Verify each owner account against the procurement record where possible (the bid was submitted, the payment was late, the criterion applied), and retain all transcripts. The 2021 study's anecdotal record (informal networks, repeat firms, unmonitored good-faith efforts, weak outreach) supplies the baseline themes; the update tests whether each theme persists, faded, or changed form after the reforms. Report corroboration by theme, and report where the current record does not corroborate a prior theme just as plainly.
Courts applying Croson require statistical disparity be corroborated by significant anecdotal evidence, and the standard attack is that anecdotes are unverified or out of context. Verified, retained, stage-coded accounts rule out the noise explanation and tell whether the experiential signature matches the statistical location, while broad recruitment rules out a self-selected-complaint bias.
Apply a dual test to every group-by-category-by-outcome cell: a substantial disparity threshold and a significance test with confidence intervals, under a pre-registered Benjamini-Hochberg false-discovery-rate correction across the full grid and a documented minimum-cell-size rule. Build the findings as a group-by-category matrix marking each cell disparity or no disparity, on the same five-category basis as the 2021 study so the update reads against the baseline. Report honest counter-results the way the 2021 study's own detail did: it found Asian American-owned firms overutilized in Construction (index 8.31, driven by the single dominant firm) and women-owned firms overutilized in Architecture & Engineering (1.76), inside the same record that measured zeros elsewhere. Pre-register the methodology and apply the test uniformly so the study can return a null.
Rules out the noise explanation (a gap from small numbers or chance) and the over-inclusion problem (treating all minorities as one undifferentiated group). The controlling Fourth Circuit case upheld measures only for the groups the evidence supported and struck the rest, so significance plus group-by-category specificity is the binding constraint that decides which findings from every prior step are real and remediable.
What it takes
The data this analysis requires
The deeper analysis is only as good as its inputs. This is the full data picture, what we collect ourselves and what the City of Frederick would provide, so the work is concrete rather than aspirational.
| Dataset | Holder | Why it is needed | Source |
|---|---|---|---|
Transaction-level payment records (both platform eras) City provides | City of Frederick Purchasing (Budget & Administration) and the financial system of record | This is the actual utilization numerator, and in Frederick it does not exist publicly: the City publishes no transaction-level spend or utilization dashboard, and the only public program figure is the 2022 review's roughly 20% of $44M. The internal extract lets HSG compute utilization by group, category, and dollar band on a consistent basis across the platform migration, which is also what makes the reform-era time series possible. It is the City's own marketplace, which Croson requires. | City financial-system accounts-payable extract for the five most recent fiscal years: every payment with vendor identifier, dollar, date, department, and commodity coding, spanning the IonWave era and the OpenGov era with a documented crosswalk across the January 1, 2026 migration, plus bridge years toward FY2019 where records exist |
Award and contract records with identifiers City provides | City of Frederick Purchasing | The 2021 study's data-reform recommendations began with formal award identifiers and tracking of awards under $50,000, because without them awards cannot be joined to payments or sequenced per firm. Requesting the award spine explicitly, and documenting where identifiers are missing, both powers the analysis and measures how far the City's data architecture has come since the study said it needed to change. | Award and contract history for the same five fiscal years: solicitation number, method, department, award date, awarded and not-to-exceed amounts, modification history, including awards under $50,000, with formal award identifiers where they exist and source fields to construct them where they do not |
Solicitation, bid, and plan-holder records City provides | City of Frederick Purchasing; platform records held with IonWave (legacy) and OpenGov | This is the dataset that separates a supply problem from a selection problem, the core of pinpointing causality. If diverse firms are available but rarely bid, the cause is upstream (awareness, bundling, bonding); if they bid at expected rates and lose disproportionately, the cause is in evaluation and selection. The 2021 study's zero-utilization categories make this decisive for Frederick: the fix depends entirely on which side of the bid stage the zero comes from. | Bid tabulations, quote logs, responsiveness determinations, and plan-holder or document-download lists per solicitation, from both IonWave and OpenGov, listing every firm that participated, not only winners |
Subcontractor utilization and payment records City provides | City of Frederick Contract Compliance Administrator (Economic Development) | Disparity often concentrates at the subcontract level, and Frederick's record makes this the highest-stakes request in the study: the 2021 study found zero tracked subcontractors in FY2018, and Frederick County's 2026 study could not support a race-conscious program because only 27% of prime contractors had kept subcontractor records. Whatever the current coverage turns out to be, it is either the evidence base for the subcontract analysis or a documented finding with a remediation plan attached. Both outcomes serve the City. | Subcontractor utilization plans submitted at bid, post-award subcontractor payment confirmations, and good-faith-effort documentation on construction contracts above $250,000 carrying MWBE goals, with a coverage statement by fiscal year and department |
SBR, MWBE, and program-administration records City provides | City of Frederick Contract Compliance Administrator; Purchasing; the DEI Advisory Board record | The draft scope's Task 3 is an assessment of the City's supplier diversity programs, and these are the programs' own administrative records: who certified and by which of the four routes, which solicitations were reserved and what happened, which construction goals were set, met, or waived, and what left competition through the 60-plus exemptions a year. The yearly effectiveness reviews the 2021 amendments require are the program's self-evaluation record; only the 2022 edition is public, and obtaining or reconstructing the later years is itself part of the assessment. | The SBE certification roster with dates and certification paths; SBR designation records per solicitation with band and outcome; annual aspirational goal-setting records; contract-specific goals, attainment, and waivers; the exemption log; the 2022 annual effectiveness review and any later reviews; DEI Advisory Board procurement-review materials |
Vendor payment timing records City provides | City of Frederick finance function | Slow payment is one of the most consistently reported barriers in disparity studies and a distinct funnel drop-off: a firm can win work and still be squeezed out by cash-flow strain that hits undercapitalized firms hardest. Payment-timing data lets HSG test whether payment lag differs by firm ownership and connects the funnel to the access-to-credit finding rather than leaving payment as an anecdote. | Accounts-payable records with invoice date, payment date, and amount for prime and, where captured, subcontractor payments |
Purchasing-policy and solicitation-template corpus City provides | City of Frederick Purchasing | A major City-controllable cause lives in the rules themselves. Coding solicitations for experience floors, reference counts, bonding thresholds, and bundling lets HSG quantify how often facially neutral criteria mechanically exclude later-entering firms, and the policy corpus is where the SBR band discrepancy (the 2026 form versus the City's other published descriptions) gets resolved on the record rather than left ambiguous. | The Purchasing Policies in every version in force during the study period, including the November and December 2021 amendments; procurement-method definitions with thresholds and discretion levels; standing solicitation templates and standard terms; bonding and insurance schedules; every version of the SBR and MWBE program forms and guidance |
Vendor roll, classification records, and permit business names City provides | City of Frederick Purchasing (vendor file); City permitting function (permits) | Registration is the first funnel stage, and the vendor file defines who the City can currently see. The permit records answer the opposite question: which firms demonstrably work in Frederick but never enter the procurement system at all. Putting business names on permits was the 2021 study's own recommendation, and the permit-derived roster is both an availability cross-check and a ready-made outreach list the City can act on immediately. | The vendor file with reported ownership demographics and self-identified flags (400-plus vendors classified, 223 self-identified small/minority/women-owned as of the 2022 review), noting self-reported versus verified fields; plus business names on building and trade permits issued during the study period |
Custom availability survey (HSG-fielded) We collect or derive | House Strategies Group LLC (PhD-led survey team; fielded by HSG, not held by the City) | This is the single biggest upgrade over any public-data read, which must lean on equal-weight ABS firm counts. A sampled, registration-anchored custom census measures willingness and market-area presence, and deliberately declines to capacity-discount on the ground that current capacity is itself partly a product of past conditions, consistent with Croson's qualified-willing-and-able standard. Reconciling it to the 2021 study's estimates means the update measures change on a consistent basis, and it gives the aspirational-goals program the current availability base it has lacked since its numbers were fixed to FY2014-FY2018 data. | Stratified random survey of firms in the spend-derived market area, weighted to the City's category mix, telephone plus email, to a 95% confidence interval per subsector; measures willing-and-able status without capacity-discounting and reconciles to the 2021 study's availability estimates for change-over-time comparison |
U.S. Census ABS / NES-D firm data We collect or derive | U.S. Census Bureau (public bulk files; HSG retrieves and derives) | ABS and NES-D establish the market-wide picture on public, reproducible data: the growing minority- and women-owned firm base, the receipts gaps, and the sharp difference between the employer universe and the far more diverse nonemployer universe, which bears directly on how the SBR's small-purchase band should be designed. Because the files are public, the figures cannot be attacked as consultant-manufactured, and every portal number traces to its source. | Annual Business Survey 2022 and 2017 and Nonemployer Statistics by Demographics 2022: firm counts and receipts by owner sex, race, ethnicity, veteran status, industry, and geography, at the county, metro, and CSA levels |
ACS PUMS microdata (but-for regressions) We collect or derive | U.S. Census Bureau (public microdata; HSG retrieves and models) | PUMS is the marketplace evidence layer. Regressions on wages, owner earnings, and business formation show whether, holding human capital and capital-access controls constant, minority and women earners and owners still trail across the regional economy. A gap that survives those controls points to conditions bigger than City procurement, which under Croson is context for the compelling interest while keeping the City's own record, not the marketplace pattern, as the basis for any remedy. | American Community Survey Public Use Microdata Sample, multi-year, for the regional market area: wages, business-owner earnings, self-employment status, plus controls (education, age, marital status, English proficiency, disability, homeownership and home value, mortgage, unearned income) |
Access-to-credit and capital-access data We collect or derive | Public sources (FFIEC, Federal Reserve, SBA); HSG compiles and analyzes | Capital constraint is a leading non-City explanation the study must test rather than assume, and a major driver of why diverse firms enter later, smaller, and with thinner balance sheets. The county receipts gaps (all firms $5.256M average against $1.677M minority-owned and $886K women-owned) demand a mechanism, and the credit channel is the first candidate. Testing it both strengthens the marketplace finding and protects fairness: it locates part of the gap in economy-wide capital access rather than City fault. | HMDA mortgage and home-equity data, Federal Reserve Small Business Credit Survey, SBA lending and Surety Bond Guarantee data, and the capital proxies embedded in the PUMS controls |
Anecdotal and interview record (qualitative corpus) We collect or derive | House Strategies Group LLC (designed by the academic bench, executed by the field team; HSG holds the corpus) | Courts applying Croson require that statistical disparity be corroborated by significant anecdotal evidence, and they have devalued unverified accounts, so the protocol verifies owner accounts against the procurement record and retains transcripts. Frederick has a baseline here too: the 2021 study documented informal-network, repeat-firm, and weak-outreach themes, and the update tests which themes persist after five years of reform, which is evaluation evidence no statistics can supply alone. | Structured in-depth interviews with distinct guides for businesses and for professional and trade organizations, open-ended survey responses, and public engagement sessions, recruited broadly across Frederick's contracting community, with a verification step and full transcript retention |
Relevant geographic and product market definitions We collect or derive | Derived by HSG from City data (the underlying records are City-held; the derivation is HSG's) | Croson requires the market be the jurisdiction's own. The 2021 study drew its Washington-Baltimore-Arlington market area to capture at least 75% of prime award dollars; the update re-derives the market from where the City's dollars actually went in the new window and validates or revises that definition rather than inheriting it. That defeats the gerrymandered-market attack and makes the availability denominator match what the City actually purchases. | Derived from the City's payment and award records: the category and commodity mix the City actually buys and the geographic distribution of where its dollars go; used to set the market area and product market rather than adopting either off the shelf |
Items tagged “City provides” are the contract, payment, solicitation, bid, and program records only the City of Frederick holds. Everything else we collect or derive from public sources and our own survey.
The payoff
From cause to action
Pinpointing the cause is what makes a recommendation targeted instead of blunt. Each cause the analysis can isolate maps to a specific fix and a specific City lever. We lead with race-neutral, opportunity-focused measures and reserve anything race-conscious for the narrow, significant, group-specific gaps that neutral fixes cannot close.
Drop-off concentrated at the vendor-registration and certification stage (ready firms never enter the City's vendor file or any of the four certification routes, so the supply of bid-eligible diverse firms is thin before any solicitation runs)
Run targeted registration drives built from the permit-derived business roster and regional business associations, simplify OpenGov vendor onboarding, and streamline the four certification routes (SBA, MDOT, USDOT, City self-certification) so one submission clears the City's requirement, with certification turnaround measured and published
Certification streamlining and vendor-registration outreach through the Contract Compliance Administrator
Race-neutralDrop-off at the awareness stage (registered diverse firms are in the pool but do not learn of relevant solicitations in time to respond, and quote lists skew to incumbents)
Extend the City's monthly procurement forecast page into a rolling 12-to-18-month forecast, push targeted notices to registered firms by commodity code through OpenGov, and hold pre-bid sessions timed early enough to prepare a bid
The existing forecast page and OpenGov notification capability, pointed at the firms the data says are missing
Race-neutralExperience floors and prior-similar-project minimums in solicitations exceed what the scope actually requires, screening out capable firms at responsibility review before price is read
Audit and right-size past-performance language to the work, accept comparable, aggregated, or subcontractor experience and key-personnel experience in place of firm-level history, and document the justification for any threshold retained
Solicitation drafting and qualification-setting standards in the Purchasing Policies
Race-neutralDrop-off concentrated at the bidding stage for construction, where bonding capacity caps who can bid as a prime rather than firm capability
Stand up a bonding-assistance track (threshold scaling on smaller jobs, a referral partnership with the SBA Surety Bond Guarantee Program, and surety-readiness support), available to all small firms that cannot bond the full contract
Bonding assistance and solicitation bond-threshold scaling
Race-neutralLarge bundled solicitations consolidate scopes that could be procured separately, confining smaller firms to subcontract roles and foreclosing the prime path
Unbundle where the consolidation is not operationally justified, break large buys into right-sized lots, and require a written bundling justification before aggregating requirements
Unbundling and contract-structuring policy
Race-neutralAwards concentrated among repeat vendors reached through informal referral, small-purchase discretion, and the exemption lane (the informal-network pattern the 2021 study documented, operating through method choice rather than any formal rule)
Move informal opportunities onto the open record by advertising small-purchase and quote-based buys, rotating quote solicitations across the registered and certified pool, tightening exemption justifications and reporting, and instituting structured matchmaking that introduces unaffiliated firms to buyers and primes
Quote rotation, open advertising of informal buys, exemption-log discipline, and matchmaking
Race-neutralSlow payment strains cash flow so that undercapitalized firms decline larger awards or cannot carry the gap between performance and payment
Enforce and tighten prompt-payment timelines, add progress-payment acceleration and mobilization support on larger contracts, and require primes to flow prompt payment down to subcontractors with monitored compliance
Prompt-payment policy and progress-payment terms
Race-neutralSubcontract participation lags because goals and good-faith efforts are set but not monitored, allowing bid shopping and after-award substitution of listed subcontractors (the unmonitored-good-faith-efforts finding in the City's own 2021 record)
Require subcontractor listing at bid on goal-carrying contracts, verify good-faith effort against a published standard, monitor actual subcontractor payments against the plan, and bar post-award substitution without City approval
Good-faith-effort enforcement and subcontractor payment monitoring under the Contract Compliance Administrator
Race-neutralRestrictive or proprietary specifications (brand-name calls, narrow product specs, or credential requirements) limit the field for reasons unrelated to performance
Convert proprietary specs to performance-based or or-equal language, scrutinize each restrictive criterion for necessity, and require justification for any single-source specification
Solicitation drafting and the procurement-criteria appropriateness audit
Race-neutralThe binding constraint is upstream and marketplace-wide access to capital (firms are undercapitalized at formation and cannot finance working capital), a condition that reaches well beyond City contracting
Partner with regional CDFIs and community banks on working-capital and mobilization financing referrals for small contractors, and pair the financing referral with the prompt-payment fixes that reduce the capital a contract demands; treat the capital gap as a shared-marketplace condition the City mitigates rather than one it caused
Capital-access partnerships plus prompt-payment and progress-payment terms
Race-neutralFirms cluster in the small-purchase band and never graduate to larger work because they cannot accumulate the past performance and bonding history larger awards require (a self-perpetuating capacity ceiling, not current incapacity)
Build a graduation pathway on top of the Small Business Reserve: mentor-protege pairing with established primes, verifiable past performance on staged scopes, and step-ups in contract size as firms demonstrate readiness, so the SBR is a ramp rather than a ceiling
The SBR band plus a mentor-protege and graduation track
Race-neutralThe evidence base itself is the barrier: awards lack identifiers, commodity codes are vague, subcontractors go untracked, and no utilization reporting is published, so neither the City nor the public can see participation (the 2021 study's data-reform finding, and the failure mode the county's 2026 study demonstrated next door)
Implement the data architecture as a deliverable, not a suggestion: formal award identifiers, commodity-code hygiene, mandatory subcontractor reporting on goal-carrying contracts, tracking of awards under $50,000, business names on permits, and a public utilization dashboard built on the OpenGov platform the City already owns, feeding the annual effectiveness reviews the 2021 policies already require
OpenGov configuration, the annual-review requirement, and permit-system fields
Race-neutralA measured gap traces to a legitimate non-discriminatory difference (for example, diverse firms in a category are concentrated in lines of work the City rarely buys, or differ systematically in firm size matched to the work), and the gap does not survive controls for those factors
Make no remedy on this finding, document the non-discriminatory explanation, and redirect effort to categories where a controlled gap remains; revisit if the marketplace composition shifts
None required; route engagement effort to general supplier-diversity outreach where the data warrants
Race-neutralFor a given group and category the index falls below the substantial-disparity threshold but the gap does not survive the dual test (it is not statistically significant once availability is restricted to qualified, willing, and able firms and controls are applied)
Record no disparity finding for that group-and-category cell, set no goal there, and route the firms into the race-neutral SBR track rather than a group-specific measure
Race-neutral SBR track; no goal-setting for that cell
Race-neutralAfter race-neutral measures are documented as insufficient to close it, a statistically significant, group-and-category-specific disparity persists and is corroborated by verified anecdotal evidence that neutral measures cannot reach
Set a narrowly tailored, group-and-category-specific participation goal scaled to the proven shortfall, built severable so no element's invalidation reaches the rest of the program, sunset-dated with a fixed review cycle tied to the annual effectiveness reviews, and retired once parity holds; apply it only to the groups and categories the evidence supports, the discipline the Fourth Circuit's controlling case enforces
Group-specific goal-setting under the MWBE program, time-limited, severable, and evidence-bounded
Race-conscious, if warrantedHow we keep it fair and defensible
The commitments that keep the analysis honest
A nuanced study introduces ways to overreach. These are the guardrails we build in from the start, several of them required by Croson and the Fourth Circuit standard that controls in Maryland, so the work is fair to every firm and able to withstand the most skeptical review.
See it in the rest of the portal
The disparity index is the surface read. The barriers module is the catalog of causes this approach tests. The methodology page draws the line between what public data can show and what the full study adds.
Source: Approach framework generated 2026-08-16.