The Gap Between the Plan and the Nail

A nail that misses the truss it was aimed at holds nothing to nothing. It goes through the plywood sheathing and into air. From the street the roof looks finished, and on the inspection card the sheathing stage is marked complete. The nail is there. It just isn't attached to anything.

In December 1992 the Miami Herald sent engineer Eugenio Santiago into four houses in Country Walk, a subdivision in south Miami-Dade County where Hurricane Andrew had come through four months earlier. Nothing Santiago found was exotic. Gable-end bracing, the structural members that keep the triangular wall at the peak of a roof from folding outward when the wind leans on it, was weak where it existed and absent where it didn't. The braces that had been installed were poorly attached. Rows of fasteners meant to tie the roof sheathing to the trusses underneath had missed the trusses.

Country Walk sat in the path of Andrew's worst winds when the storm crossed the coast on August 24, 1992. Arvida had built roughly 1,300 of the development's approximately 1,700 homes. The Herald's analysis of some 60,000 post-storm damage records found that at least 90 percent of inspected homes in the broader Country Walk area were ruled uninhabitable. That classification swept in houses needing major repair along with the ones reduced to slabs and scattered lumber. "Uninhabitable" is an administrative finding about whether a person can live in the place now. It says nothing about what failed or why, which hasn't stopped it from being read that way ever since.

Country Walk mattered to the investigation that followed because of what turned up when people began asking why some houses came apart while others, built under the same code in the same years in the same wind, did not.

A Third of a Mile

About a third of a mile south of Country Walk sat Munne Estates: 71 houses built in 1989 and 1990 with concrete-block walls, plywood sheathing thicker than the code minimum, and hand-driven nails instead of pneumatic staples. Nearly all the roofs stayed on.

Nearby, the numbers ran the other direction. All 38 houses in Roger Homes were declared uninhabitable. In Mediterranea, 99 percent of 111 homes got the same classification.

These were not identical developments. Munne used block where others used wood framing. Its roof geometry was simpler, its sheathing thicker, its fasteners driven by hand. And wind intensity varied across short distances; Dade County's chief building inspector, Roberto Pineiro, argued at the time that the wind in certain spots was enough to tear up sound construction no matter how it was built.

Building 1,300 houses in one development means pneumatic staple guns instead of hammers, prefabricated trusses set by crane, crews moving lot to lot on a schedule somebody wrote down in an office. A staple gun fires the same whether there's a truss behind the plywood or nothing behind it; a man swinging a hammer feels the difference. The 1989 Dade County grand jury had already noted that workmanship across the county was widely condemned and that roof work was being done by general contractors rather than licensed roofing contractors. Munne's builder went the other way: block and thicker plywood and hand nailing, which cost more and took longer. Why one builder made those choices and another didn't is a question the public record mostly leaves alone. The investigations after Andrew documented what was in the rubble and in the inspection files. The production side — who was driving the fasteners and how they were trained and what schedule they were on and who was watching the crew that morning — is known almost entirely by its results.

The Munne comparison is not a controlled experiment and nobody should pretend it is. What it establishes is that near-total destruction was not the uniform, inevitable outcome of a Category 5 hurricane crossing recently built houses. Some of those houses stood.

What the Rubble Showed

The FEMA Building Performance Assessment Team surveyed damaged and undamaged residential buildings across south Dade in the weeks after the storm. Its December 1992 report found failures at nearly every connection point in what engineers call the load path: the sequence of fastened joints that carries wind force from the roof down through the walls and into the foundation. Every joint has to hold or the sequence breaks. FEMA found them failing at every stage.

Staples that missed the top chord of the truss, the horizontal member they needed to bite into. Sheathing the report described as "only tack-nailed." Hurricane straps, which tie roof trusses to the top of the wall, misplaced or missing. At one house FEMA photographed a sill plate, the bottom member of a wall where it meets the foundation, fastened with cut nails and no bolts. A bolt threads into concrete and holds against a pull. A cut nail in a mortar joint slides out. The wall was sitting on the foundation more than it was attached to it. Elsewhere: masonry with poor mortar joints, missing tie beams (the reinforced concrete band that caps a block wall and gives the roof something to fasten to), absent reinforcement, shingles held by too few staples or by staples driven sideways.

The report's photographs generally don't say which subdivision a particular failure came from. It catalogs how things broke, not where.

Structural engineer John Pistorino walked damaged neighborhoods with a Washington Post reporter shortly after the storm. At one house he found hurricane straps that had been installed and never attached, the metal connectors sitting in place with no nails in them.

At another stop he pointed to a collapsed gable roof and described a failure that ran back through the design documents, the code's allocation of responsibility, and inspections that had checked the construction against incomplete plans.

Haag engineer Timothy Marshall, who ran independent aerial and ground surveys the week after Andrew, documented trusses that had rotated because the lateral bridging that keeps them from twisting wasn't there. He photographed a whole sheathing panel that left with its shingles still attached, never having been properly fastened to the trusses beneath it. Gable ends that fell outward because nobody tied them into the roof framing. Hurricane straps nailed to the face of a truss rather than wrapped over the top of it. A strap on the face resists the truss sliding sideways; over the top, it resists the truss lifting, which was what the wind was asking for.

Mohammed S. Khan and Wimal Suaris did the most systematic subdivision-level work, examining seven south Dade developments, Country Walk and Lakes by the Bay among them, for the Miami-Dade Building Department and the County Manager's code task force. Their published findings reported masonry reinforcement missing or improperly anchored, wood gable ends unbraced, roof sheathing inadequately fastened. The underlying work ran to ten volumes under the title Identified Code Violations and Construction Deficiencies. Its house-level data has not been widely accessible since.

What these investigators were describing, between them, was a set of structural connections that existed on paper and not in the buildings. The code called for a continuous path from roof to foundation. The plans were supposed to show how it would be made. The inspections were supposed to confirm that what got built matched what got approved. Each stage took the previous one on faith.

What the 1989 Grand Jury Already Knew

None of this was news. Three years before Andrew a Dade County grand jury had looked at the Building and Zoning Department and found an operation running well past what it could handle.

The department's own performance study reported 28.7 inspections per inspector per day, averaging nine minutes apiece. Outside experts testified that an inspection ought to take closer to fifteen minutes and that the daily number ought to be lower. Formal quotas had gone from 22 a day in 1985 to 27 in 1989 and 29 in 1990.

The grand jury didn't stop at the statistics. It put inspectors under surveillance. One man's site visits ran between one and ten minutes, half of them between one and four, and he went home at 12:05 p.m. Another listed ten inspections at three stops, never approached the equipment at one of them, never visited the third at all, and spent part of the afternoon in a library reading newspapers and sleeping.

Roofing was a particular blind spot. The surveillance team never saw a roofing inspector use a ladder. And the county's inspection sequence had a gap you could drive a truck through: nothing was required between the tin-cap stage, when the basic roof frame is up, and the final inspection. Everything that happened to a roof in between got covered over before anybody looked at it.

The grand jury recommended realistic workloads, more inspectors, quality assurance, removal of employees who falsified records, continuing training. The Herald later found that the county had issued 194 daily assignments since 1987 carrying more than fifty inspections apiece, better than twice the twenty-a-day maximum the grand jury had asked for.

So the department had been audited, the audit found it broken, the recommendations were written down, and the quota went up. Three years later the hurricane came.

Products the Code Allowed

Field inspection was not the only weak link. Some of what failed during Andrew was permitted outright by the South Florida Building Code.

Dade County's Board of Rules and Appeals had allowed staples as a substitute for nails in roof construction since 1961. By 1983 the board had been warned that staples performed badly in lesser tropical events. The provision stayed on the books until after Andrew. Because staples were permitted, their presence wasn't a violation; the violations were in the particulars — staples that missed framing, staples driven in the wrong orientation, too few of them, spaced too far apart for the loads involved.

The code also permitted asphalt shingles rated to resist roughly 63 mph in a jurisdiction whose structural wind benchmark was 120 mph. The board had approved waferboard as a plywood substitute in 1984, over reported concerns inside Dade's product-control division about industry-sponsored testing; waferboard was banned after Andrew because it came apart when it got wet. Thinner roofing felt had been approved in 1980 to save money.

The post-Andrew grand jury found a problem built into the approval process itself. Independent laboratories conducted the product tests, and the manufacturer paid the laboratory. The grand jury recommended that the county collect the fees and pay the labs directly. You don't need an engineering degree to see why the man writing the check shouldn't be picking the man who grades the test.

These are different kinds of failure: a permitted fastener unforgiving of installation error, an approved covering rated for about half the design wind speed, a substitute sheathing whose wet performance raised concerns nobody acted on. The code was a system of approvals, specifications, substitutions, installations, and inspections, and it could break at any one of them.

Systematic Failure

The first post-Andrew grand jury, reporting in the fall of 1992, concluded that the storm's unavoidable destruction had been compounded by "systematic failure" in construction and building regulation. Its follow-up report called the lack of proper inspections a "major contributing factor" and said the storm had exposed faulty construction that had been hidden until then.

The diagnosis went past inspection. Plan review had not ensured that tract-development designs provided the load transfer the code required. Plans treated roof sheathing as the only permanent bracing for prefabricated trusses, which made the thing holding the roof structure together the same thing the wind was trying to peel off. The grand jury named inadequate design, poor workmanship, deficient product approvals, and missing inspections as connected problems, and it put workmanship alongside enforcement rather than treating the builders as bystanders in somebody else's failure.

After Andrew, Dade County went from 16 building inspectors and supervisors to 43, and from 4 roofing inspectors and field supervisors to 31. Officials told the grand jury that inspectors were still averaging 22 to 25 inspections a day. The grand jury recommended adding roof bracing to the newly created sheathing-inspection stage, attaching the failure Andrew had exposed to a moment in construction when somebody could still see it.

Four years later the Governor's Building Codes Study Commission delivered the broader verdict. Florida's building-code system was "protective of the health, safety and welfare" in principle but "ineffective in its execution," particularly in administration, enforcement, plan review, inspection, training, discipline, and industry compliance. The commission identified three connected deficiencies: the number and inconsistency of codes across the state's more than 400 jurisdictions; the administration and enforcement of those codes; and compliance by the design and construction industries.

The Florida Building Commission's later institutional history put it more simply: the problem in Andrew was not weakness in the codes but an inability to enforce and comply with a confusing system administered differently across hundreds of jurisdictions. That's a retrospective compression. The contemporaneous grand jury record shows code provisions, product approvals, plan review, and field inspection all contributing. But even the study commission's own language treated the written technical rules as one element of a system that also had to approve plans, evaluate products, train the people doing the work, inspect what they did, and discipline the ones who cut corners.

One Code

The 1998 Florida Legislature enacted House Bill 4181, adopting the study commission's central recommendation: one statewide building code administered by a Florida Building Commission, with statewide product evaluation and approval, education and training requirements, clearer lines of accountability, and local-government enforcement of the unified rules. The first Florida Building Code took effect on March 1, 2002, nearly ten years after Andrew.

Local building departments still did the permitting, the plan review, and the inspections. The reform unified the base code and the systems around it. It did not replace field enforcement with a state inspection force. Whether the new arrangement would hold under comparable stress stayed an open question until hurricanes tested it.

What the Record Holds

A few things about the limits of this evidence should be said plainly.

The FEMA photographs of missed staples, unattached straps, and failed connections are not consistently tagged to particular subdivisions. Santiago's four houses are the most geographically precise construction evidence I've located for Country Walk itself. The often-repeated figure that 90 percent of Country Walk was "destroyed" isn't supported by the strongest contemporary data; the Herald's December 20, 1992 "What Went Wrong" analysis found 90 percent uninhabitable across the broader area, a category that included repairable damage. The ten-volume Khan study of seven subdivisions would likely hold the most detailed address-level evidence that exists, and it has not been widely accessible.

The Munne Estates comparison, 71 houses with their roofs on a third of a mile from near-total destruction, is powerful and uncontrolled. Construction type, roof geometry, sheathing thickness, fastener choice, builder practice, and local wind exposure all varied. What it establishes is that the outcome wasn't uniform, and that something besides wind speed determined which houses stood.

In an earlier piece for this publication I worked through the different mortality counts that came out of Chicago's 1995 heat disaster and argued that the disagreement among them showed how a disaster becomes administratively visible, rather than describing one fixed number. The same applies here. "Destroyed," "uninhabitable," "damaged," and "code-compliant" are administrative findings with different definitions and different denominators. Collapsing them into a single number makes the story easier to tell and the evidence less honest.

What the record does establish is that the difference between a house that stood and a house that came apart was, over and over, at the level of whether a fastener met the wood it was aimed at. That gap was on the books before the storm, visible in the rubble after it, and named afterward as the central systemic failure. The remedy — one code with standardized product approval and statewide training — took ten years to enact and left the hardest part where it had always been, with local building departments working through the day's assignments.

The 1989 grand jury had run the same audit without the wind. The county's answer was to raise the daily quota to 29.