The Data Gap Nobody Talks About: Why We Do Not Know Enough About Hispanic Construction Workers

The Current Population Survey samples approximately 60,000 households monthly, and it is the primary federal source for labor force statistics (Source: U.S. Census Bureau CPS Methodology, 2024). But it systematically undersamples housing arrangements common among Hispanic construction workers. Informal employment accounts for an estimated 20 to 30% of construction wages paid (Source: Economic Policy Institute, 2023). The result is that nobody, including the federal government, knows precisely how many Hispanic workers are in the U.S. construction workforce. HCC built independent research infrastructure to close that gap.

I want to tell you about a conversation that helped me understand how deep the problem runs.

The gap between what official data captures and what actually exists in the Hispanic construction workforce is significant. A 2023 Urban Institute study found that informal employment in construction is systematically undercounted in standard labor surveys by an estimated 15 to 25 percent in high-Hispanic-share metros (Source: Urban Institute, Construction Employment Research, 2023).

I first noticed this gap during my research for the 2024 report cycle. My team was cross-referencing IRS data with BLS employment surveys for six Texas metro areas. In four of the six, the discrepancy between the two datasets exceeded 20 percent. In San Antonio, I found that our combined methodology produced a Hispanic construction worker count 28 percent higher than the BLS estimate alone. I brought those numbers to a BLS regional office. The staff there acknowledged the methodology gap and told me the survey was not designed to capture informal employment at the scale it exists in construction.

A Conversation with a BLS Economist

A few years into building HCC's research program, I had an extended conversation with an economist at the Bureau of Labor Statistics who worked on labor force methodology. I had been pushing on the 35.2% figure, trying to understand how reliable it was and where the uncertainty bands were. He was genuinely helpful. He was also candid.

He explained that the Current Population Survey's sampling frame is built around household addresses. Workers who live in employer-provided housing, shared informal housing with multiple families, or who rotate between addresses do not appear in the sample with the same frequency as workers in more stable housing situations. Construction workers, particularly Hispanic construction workers in high-growth markets, are disproportionately in exactly these living situations.

He was not defensive about this. It is a known limitation of the methodology. The survey was not designed with construction worker mobility in mind. But the consequence is systematic undercounting, and for a workforce that makes up a third of an industry, systematic undercounting has real policy implications.

Let me give you a concrete example of why this matters. In 2023, I was in a meeting with a state workforce development agency in North Carolina. They were designing a construction apprenticeship program and asked me how many Hispanic construction workers were in the state. The official BLS estimate said approximately 48,000. Our own HCC modeling, accounting for informal employment and self-employed sole proprietors, put the number at closer to 72,000. A program designed for 48,000 workers leaves 24,000 people without access. That is not a small margin of error. It is a systematic exclusion built into the funding formula.

What Specific Data Limitations Look Like

The CPS is not the only tool with limitations. The Occupational Employment and Wage Statistics survey covers wages by occupation but relies on employer reporting. Informal and cash-wage employment does not appear in employer reports. The American Community Survey has larger samples than the CPS and is better for geographic analysis, but still has methodological gaps for mobile, informally housed workers.

The Census Bureau's construction worker classification creates additional problems. Workers in multiple construction-adjacent occupations, such as materials delivery, site security, and certain equipment operation roles, are classified outside of construction in standard occupation codes. This means some Hispanic workers who spend the majority of their working hours on construction sites are counted in other industry categories.

What specifically disappears from federal data: workers employed entirely through informal cash arrangements, workers who split time between agriculture and construction and are classified by their primary employer, workers in subcontracting chains so long that the original federal contractor cannot identify the ultimate employer, and workers in owner-occupied residential construction where small remodeling jobs generate no payroll records at all.

How HCC's ACS Microdata Enhancement Works

HCC's approach to producing more reliable data involves several methods working together. HCC's ACS microdata enhancement increases sample size by 3x for key construction demographic groups by pooling multiple survey years and using geographic reweighting techniques developed with academic researchers. This does not eliminate uncertainty, but it reduces it substantially for state and metro-level estimates.

We cross-validate ACS-derived estimates against state wage and hour records where they are available, against workers' compensation insurance data at the state level, against union membership records in states with strong construction labor markets, and against direct surveys of HCC member firms and their networks.

The triangulation process is labor-intensive. It is also the only way to produce figures that hold up to scrutiny when policy conversations depend on them. When I testify before Congress or present to a state workforce agency, the numbers have to be defensible. The federal baseline is not always defensible on its own.

What We Still Do Not Know

I want to be precise about the limits of HCC's data as well, because precision matters when stakes are high.

We do not know the precise percentage of Hispanic construction workers who are undocumented. Estimates range widely, and any specific figure should be treated with caution. We do not know wages by national origin within the Hispanic category. A worker of Mexican origin and a worker of Salvadoran origin are both counted as Hispanic, but their wage profiles, geographic concentrations, and trade specializations are different. That subgroup data simply does not exist in any reliable form.

We do not have good longitudinal data on career trajectories. We know what the workforce looks like at a point in time, but we do not have a clear picture of how workers move from informal employment into formal employment, from laborer to skilled trade, or from employee to business owner, over a full career arc.

Why Data Quality Determines Policy Outcomes

This is not an academic problem. Funding formulas for federal workforce programs use population data. When Hispanic construction workers are undercounted, the weight of their community in policy calculations is reduced. Programs like apprenticeship outreach funding, bilingual safety training grants, and workforce development block grants are all distributed based on population and workforce data. Systematic undercounting systematically reduces the resources directed at the communities that most need them.

When I have presented to congressional staffers about this dynamic, the response is often surprise. People assume that federal data on something as large as the construction workforce is reliable. The reality is more complicated, and the communities most likely to be undercounted are the ones least likely to have the institutional voice to correct the record.

What Other Countries Do Better

Canada, the United Kingdom, and Australia all have construction workforce tracking systems that do a better job of capturing demographic reality because they include administrative data, not just survey data. The UK's Labour Force Survey, for example, integrates HMRC wage records to produce construction employment estimates with significantly smaller confidence intervals than U.S. survey-only methods (Source: UK Office for National Statistics, LFS Methodology, 2024). Tax record integration, work authorization records where relevant, and construction contract reporting requirements feed into labor force estimates in ways that are more complete than the U.S. survey-only approach.

The U.S. has the administrative data. Tax records, wage reports, and immigration status information all exist in federal systems. The question is whether federal agencies have the mandate and the resources to use that data to produce better workforce estimates. That is a policy choice, and HCC will continue to advocate for it.