A building's function always includes the provision of a safe and stable platform—the building's "structure"—to enable its other functions. Structure—describing the elements that support and define floors, roofs, and exterior walls—can be more or less efficient (e.g., measured by the weight of structural steel required for each square foot of floor area), can interfere with or facilitate building activities (e.g., columns in an auditorium or stadium that block sight lines from certain seats), can support or inhibit future changes in occupancy (flexibility), and can be more or less safe (although structures in advanced capitalist democracies generally achieve a fairly consistent level of safety, i.e., are designed to have an economically or politically tolerable probability of failure, consistent with the intended occupancy) .
While it may seem self-evident which fixed elements constitute a building's structure, there is actually no rational way to distinguish structural from so-called non-structural elements. Any attempt to define structure in such a way that certain solid building elements are excluded will run into problems. Defining structural systems as "load-resisting" or "transferring loads," per Wikipedia,1 is pretty much useless, since a carpet placed over a concrete slab also resists dead and live loads placed on it, and transfers those loads to the slab beneath it. Is that carpet really part of the building's structure? In the same way, a pane of glass in a window is also structure, since it resists wind loading, and transfers those wind loads to adjacent walls or to the building's structural frame. In fact, everything fixed in a building is structure since everything both resists and transfers loads, even if only its own self-weight. But such an all-inclusive definition is clearly unsatisfactory in this context. We need to limit the firmitas category by excluding carpet and similar things—those things labeled non-structural in common usage. The best way to do this is to define structure in a circular manner, as those elements or systems in a building that have been designed by a structural engineer.
This much can be said about the objective, or utilitarian, qualities of structure: now largely in the hands of engineering consultants (i.e., removed from the purview of architects), the structural function is informed by engineering science advanced through academic research and practical experience, codified in consensus-based protocols developed by not-for-profit organizations representing the major structural materials, and finally adopted within a political process as a part of actual or de facto national building codes. For this reason, the utilitarian function of structure in advanced capitalist democracies has become largely routine, employing a limited range of materials, shapes, and connections, and achieving a fairly predictable and consistent level of safety (or, put another way, a fairly consistent and politically acceptable probability of failure). This is not to deny the continual refinement and advancement of structural design (e.g., improvements of structural systems designed to resist seismic or wind loading) or the combination of routine structural elements into creatively outlandish systems consistent with the architectural forms they support, but only to emphasize the rational nature of such advancements or applications when examined solely from the functional standpoint of firmitas: providing adequate strength, elasticity, and stiffness.
The utilitarian structural function also has a political dimension since the design of all structural systems—based on government-sanctioned standards—presupposes a politically acceptable probability of failure. There are two aspects to this observation. First, it is not possible to create standards for structural systems that absolutely preclude failure. Uncertainties in three categories discussed below—that is, concerning the design, manufacture, and construction of structural systems—can be minimized, but never entirely overcome.
Numerical models for structural design only approximate the behavior of real structures; even the most sophisticated finite element methods, which subdivide the "real" structural system into a conceptual matrix of ever-smaller parts in order to better mathematically relate the behavior of the various parts to each other and thereby determine the distribution of stress and deflection under assumed loads, face practical limits in how small these parts can be, and face theoretical limits in how well the interaction between the parts can be modeled. The question of what loads to apply is even more fraught with uncertainty; actual loads are both non-stationary and dynamic in nature, far more complex than the static force vectors typically assumed in structural calculations. And extreme loading scenarios, especially those caused by seismic events (earthquakes) or high winds (hurricanes/typhoons), are not only inherently unique and unpredictable, but also have different probabilities of occurrence and different probabilities of exceeding certain magnitudes of acceleration (ground motion) or wind speed, depending on geographic location. Dead loads—consisting of the weight of the building and its fixed elements—can be anticipated with a high, but not absolute, degree of confidence; but that assumes that the building materials specified for the design remain in place forever, and are not replaced with heavier (or lighter) material. For example, heavy ceramic tile on a mortar bed may weigh four times as much as the hardwood floor it replaces, eight times as much as carpet, and up to 16 times as much as thin linoleum or asphalt tile. And while being heavier has obvious ramifications for structural safety, being lighter than designed can also be problematic, especially for tall buildings where adequate weight, bearing down in the vertical direction (i.e., through the action of gravity), may be necessary to prevent uplift or overturning tendencies triggered by horizontal wind loads. Live loads—those caused by things like people or moveable furniture—are, by definition, unpredictable. How could one possibly know in advance not only how many people will fill any given room or any given building on any particular occasion, but also how much they will weigh, whether they will all congregate on one edge of the structure or distribute themselves evenly throughout the building, whether they will be jumping or dancing in unison, imparting dynamic loads on a structure designed on the basis of static loading, and so on.
Manufacturing processes have varying degrees of quality control. Wood elements, at one extreme, have huge variations in physical properties that cannot always be accurately determined after they are cut from trees and evaluated (graded); concrete strength, especially when elements are cast in place on site, has such a high potential for uncertainty that numerous samples must be shipped to testing labs and cured for four weeks before being crushed to confirm whether the specified design strength has, in fact, been met. Such tests may provide incentives for contractors and suppliers not to cut corners, but also may provide incentives for unscrupulous testing labs to fabricate results in order to save money.2 Even the steel industry has had its share of scandals, for example, the disclosure by Kobe Steel "that some of its executives had known about fake quality data for years—in at least one case for decades," a problem attributed to "a companywide focus on profitability and weak corporate governance."3
Construction offers a final dose of uncertainty. Even if manufacture and design seem adequate, construction of the building's structure—involving innumerable acts of cutting, placing, adhering, connecting, and so on—may not conform to the design drawings and specifications. Wood elements may not be nailed or bolted as specified, or they may suffer unintended damage due to rain; adhesives may be applied to surfaces that are no longer clean, or used when the ambient temperature falls outside the manufacturer's specifications. Concrete reinforcement detailing may not correspond to the design drawings, or so-called honeycombing may result from improper consolidation techniques—discovered only after formwork is removed. Steel connections may not be properly executed: welds may be too porous or have the wrong bead profile. High-strength bolts may not be tightened to deliver the correct tension; and so on.
This is the bad news: many things can go wrong in the design, manufacture, and construction of structural systems. The good news is that structural design explicitly acknowledges such uncertainties, which brings up the second aspect to the observation that the utilitarian structural function has a political content. Once it is understood that structural failure is impossible to prevent absolutely, the following question arises: how safe should structures be? Unfortunately, any seemingly rational and objective mathematically based methodology that might possibly shed light on this question ultimately falls apart when confronting the only common metric recognized as valid within a free-market society—cost. Even though we might agree with J.G. MacGregor, the internationally renowned engineering researcher who argued in 1976 that the function of structural safety must account for the fact that "(a) the strengths of materials or elements may be less than expected, (b) overloads may occur, and (c) the consequences of a failure may be very severe,"4 we must also acknowledge that none of these three issues can be addressed from a purely objective standpoint.
Skipping over, for a moment, strengths of materials and magnitudes of loads, the third issue—measuring and evaluating the "consequences of failure"—presents a unique challenge. Since the consequences—the "cost"—of injury and loss of life involve considerations of value that cannot simply be equated with the exchange value of ordinary commodities bought and sold in the marketplace, academics must twist and turn (and they do; see the discussion of fire safety in Chapter 3) to balance cost savings associated with decreased safety standards against the increased cost required to better protect against injury and loss of life. Yet while one finds plenty of academics willing to make such calculations and validate such numbers (including the value of a human life), final decisions impacting the cost of structures within a competitive marketplace are necessarily political, and so are made within legislative bodies.
In the U.S., at least since the consolidation of the major not-for-profit model code agencies under the International Code Council (ICC) and the issuance of the inaugural International Building Code in 2000, all 50 states have decided to forego competition on the basis of structural safety by agreeing to adopt the recommendations of the ICC, embodied in the ICC's suite of model codes—albeit with a few state-by-state modifications—and create a de facto national building code. This certainly does not remove the political dimension from the determination of how much safety is appropriate, since not only academics but also industry and business interests are represented in the voluntary consensus standards process through which such model codes are adopted in the first place. However, by leveling the playing field and by making any deviation from the national norm a competitive disadvantage (since businesses working across state boundaries value uniformity and certainty), the incentive for states to use decreased structural safety as a competitive bargaining chip is reduced, if not eliminated. What remains as a regulating agent, keeping both the risk of failure and the cost of construction within acceptable bounds, is the incremental accumulation of experience with structural collapse—in particular, the experience gained from ever-new terrorist, earthquake, and hurricane events—that provides not only raw data for advances in engineering knowledge but also two additional types of information that even politicians wary of imposing increased regulatory costs can comprehend: an awareness that their own lives and property, and the lives and property of the elites they represent, are in danger; and real (not merely modeled) measurements of economic loss in relation to existing standards that provide some insight into whether, and to what extent, expected future losses ought to be mitigated by adopting tougher codes.
With respect to the first two parameters listed by MacGregor— strength of materials, on the one hand, and magnitude of loads, on the other hand—the risk of structural failure is largely determined by the selection of factors of safety that effectively reduce material strength (in so-called allowable stress design) or, in more modern methodologies (strength design for reinforced concrete or load and resistance factor design for wood and steel), not only reduce the values assumed for material strengths (resistances), but also increase the values assumed for loads. Yet, since the choice of safety factors invariably affects the cost of building in relation to building safety, it, too, becomes a political question that involves exactly the same sort of considerations—about the value of life and property—raised when evaluating the consequences of failure independently of loads or material strength.
That the seemingly objective function of safety factors has a political content becomes evident when examining the historical competition between concrete and steel industry associations for market share. Many modern commercial or institutional buildings can equally well be built in steel or reinforced concrete, and the decision to use one material or the other often hinges not on esoteric questions involving structural or architectural expression, but rather on cost. Cost, in turn, is affected by the magnitude of safety factors, since safer buildings have larger, heavier, or more expensive components. Safety factors, in turn, are embedded in building codes which have been adopted through legislative processes; these codes, which are recognized as legal mandates, are based on non-binding model codes whose structural design requirements are, in turn, based on standards promulgated by the various structural material industry associations; and these competing associations—primarily the American Concrete Institute (ACI) and the American Institute of Steel Construction (AISC) in the U.S.—do not necessarily cooperate with each other when developing their own structural recommendations.
All this is illustrated by the peculiar story of how more sophisticated and explicitly risk-based methodologies were first introduced into U.S. engineering practice. Traditional use of Allowable Stress Design (ASD), with its single safety factor, neither accounts for the full (ultimate) strength of materials nor the different risks presented by different types of loads and their combinations. While it is still being used, ASD is threatened by more sophisticated structural design methodologies that—because they more explicitly account for risk—have the potential to reduce the cost of structures that might otherwise be designed with unnecessary strength (or, in some cases, to increase the cost of structures that might otherwise be designed with inadequate safety) .
Consider, for example, two structural columns, each supporting a weight of 1,000 (pounds, kips, kilograms: the units are not important here), but with column "A" supporting 750 units of live load and 250 units of dead load, and column "B" supporting 250 units of live load and 750 units of dead load. In ASD, both columns would be designed to have the same size and strength, since the total load in each case is the same. However, in a more sophisticated method, such as the strength design method used for reinforced concrete structures, separate factors of safety would be placed on live and dead loads, recognizing the higher degree of uncertainty in the specification of live loads compared with dead loads. Using strength design, since column "A" has a greater proportion of live load than column "B," it would be designed differently (and end up being stronger) than would column "B," but both columns would have the same, and an appropriate, risk of failure. In principle, then, a structure designed with ASD would be less efficient and more expensive than one based on strength design, since both ASD columns would need to be designed for the worst-case distribution of live and dead loads represented by column "A." Therefore, assuming that the ASD method was calibrated so that safety would be optimized for proportions of live and dead load found in column "A," column "B" would end up being safer, stronger, and more expensive than it needed to be. But if the calibration were based on some other assumed proportions of live and dead load—for example, if column "B" instead of column "A" turned out to be designed with an appropriate degree of safety—then column "A," having a greater proportion of live load but the same structural strength, would be cheaper, but less safe, than it should be. In other words, while there may be one "sweet spot" where the design of columns in ASD and strength design are precisely equivalent, any other proportion of live and dead loads would result either in the ASD version being more expensive (and too safe) or less expensive (but comparatively unsafe) .
The strength design method (originally called "ultimate strength design") was pioneered by the American Concrete Institute in 1956 and, as illustrated above, incorporated separate load safety factors for each different kind of load, while also considering the ultimate (failure) stress, rather than relying upon a single "allowable" stress.5 Nevertheless, an allowable ("working") stress design method remained the dominant methodology for many years because it was simpler to use. However, this latter method did not distinguish between uncertainties inherent in various load types (e.g., dead vs. live loads), and did not consider the actual (ultimate) strength of a structural element subjected to these loads.
By the early 1960s, strength design for reinforced concrete structures had matured to the point where both loads and resistances were given their own, independent sets of safety factors that were equivalent, at least in theory, to what many years later became known as load and resistance factor design (LRFD), eventually adopted by the wood and steel industries. While the traditional working stress design method was, at that time, still the featured methodology for the design of reinforced concrete elements, strength design gradually began to displace the older method. The first incarnation of strength design did not yet have explicit strength-reduction (resistance) factors and was presented somewhat tentatively in the 1956 edition of ACI 318, the "Building Code Requirements for Reinforced Concrete" that is updated by ACI every few years. A short note referred those willing to try this new method to the appendix, which contained a concise description of the requirements for "ultimate strength design." In 1963, working stress and strength methods achieved separate but equal status within the body of ACI 318. By 1971, strength design had become the featured method, with working stress design still included, but only as an "alternate design method." In 1989, working stress design no longer appeared in the main text of ACI 318 at all, but was moved to the appendix, where it remained as an alternate method for another decade. By the time ACI 318 was updated in 2002, working stress design had been consigned to a small note in the manual's commentary stating that anyone still interested in it would need to consult the appendix of the 1999 edition, where it had last appeared.
Remarkably, it took 30 years after strength design was first presented in the ACI Code before the steel industry adopted LRFD in 1986. For many years, however, load factors differed between steel and reinforced concrete. Those adopted by ACI had been calculated on the basis of "engineering judgment" rather than on more solid empirical studies and probabilistic research. Initial values from 1963, for example, included load factors of 1.5 and 1.8 for dead and live loads respectively; these were "adjusted" to 1.4 and 1.7 in 1971, where they remained for more than 30 years. Meanwhile, dead and live load factors for steel structures were set at 1.2 and 1.6 respectively, values that appeared in the very first LRFD edition of the American Institute of Steel Construction's (AISC) Manual of Steel Construction in 1986, and that have been sanctioned by the American Society of Civil Engineers (ASCE) in their Minimum Design Loads for Buildings and Other Structures since 1988 and by the American National Standards Institute (ANSI) in the precursor to this standard dating from 1982.
In principle, the lower load factors used for steel structures (1.2 and 1.6) compared with those used for reinforced concrete structures (1.4 and 1.7) would make steel structures both less expensive and less safe since they could legally be designed for smaller loads. However, because the new design methodologies contain not only load factors, but also strength-reduction safety factors affecting the assumed resistance of the structural material, the final degree of safety is determined not just by "design" loads placed on the structure, but by the combination of load and resistance factors. And while the steel industry allowed smaller loads to be used within the design process, the concrete industry was less conservative in determining the magnitude of its strength-reduction (resistance) safety factors.
Safety factors for loads ought to be completely independent of particular material properties, so it was something of an embarrassment for the concrete and steel institutes to be seen arguing in this way; and since concrete and steel elements are often used in the same building, the problem of constantly re-calculating, and potentially losing track of the magnitudes of design loads—for example, when the weight of a steel column bears on a concrete foundation pier—became not only cumbersome but dangerous. Therefore, it was something of a relief when the ACI finally gave in, reconciling their strength design load factors with those of the AISC and ASCE in the 2002 edition of ACI 318. In order to maintain a comparable level of safety with these newly reduced, and therefore less conservative, load factors, ACI 318-02 also adjusted its strength-reduction factors—that is, made them more conservative.
Thus, competition—between material-based industry groups seeking to lower the cost of their products and thereby increase market share; between individual states seeking to attract businesses on the basis of lower construction costs; and even between nations seeking to provide a more attractive "climate" for business by inadequately upgrading, or not properly enforcing, regulations that might otherwise increase costs of construction—is always a factor in the determination of structural safety. But while such competition tends to reduce the cost of construction and therefore reduce safety, the increased risk to life and property entails a countervailing cost, raising the same question we started with: how safe should structures be? Individual competitors cannot be relied upon to provide a satisfactory answer, especially in a probabilistic environment where cutting corners and reducing structural costs does not guarantee structural failure in any particular case. For example, builders constructing a reinforced concrete school in a country with ineffective, or corrupt, code enforcement may simply build cheaply and badly, especially if this is the way they are used to building, and if it is their experience that even such badly built buildings do not necessarily collapse. Everything works out just fine until, that is, a low-probability earthquake, hurricane, or tsunami strikes. In this way, individual, profit-driven decisions reveal their social/political dimension: for society as a whole, calculations about the risks to life and property can be assessed in the aggregate, and a level playing field can be established so that competition on the basis of reduced structural safety is no longer viable.
1 Wikipedia, s.v. "Structural System," last modified April 6, 2019, 04:38, https://en.wikipedia.org/wiki/Structural_system.
4 MacGregor, "Safety and Limit States," 489.
5 My discussion of steel vs. reinforced concrete safety factors is based on Ochshorn, Structural Elements, 342–43.