Milky Eggs

New York’s subway construction is strangled by government bloat

It has become increasingly well appreciated lately that “America can’t build.” More accurately, modern America seems to be incapable of building large infrastructure in anything resembling a timely, cost-effective manner, leaving us handily outdone by obscure North African countries. The prevailing diagnoses much resemble the crumbling decrepitude of human aging: the cost overruns appear to arise from a multitude of disparate inefficiencies rather than from any one key failing. I argue, however, that a new case study on the failures of New York’s Second Avenue Subway extension from the Transit Costs Project well illustrates how government bloat writ large is a unifying lens through which America’s construction inefficiencies can be productively analyzed. This analysis, in turn, helps us concretely imagine what a model of effective-but-small governance would look like.

This case study, titled The New York Case, is an in-depth, 74-page report that analyzes the sources of cost overruns in the Second Avenue Subway project’s Phase 1 construction, a 2.7 km, three-station extension of the existing subway network that racked up a bill of no less than $5.3 billion (2020 dollars), on the order of five to ten times more expensive per kilometer than some European subways:

The study is remarkably comprehensive, including detailed comparisons of stations’ floor plans against those in other countries and drawing on numerous personal interviews with contractors and government staff.

The full report comes strongly recommended, but for brevity, I summarize below the four reasons they identify as the core drivers of high construction costs for this project, many of which plausibly apply across the nation:

• Intergovernmental coordination and utilities: Active conflict and lack of cooperation between different parts of the government
• Labor wages and staffing: Over-reliance on contractors as opposed to maintaining in-house staff
• Procurement and risk: Regulations on cost overruns and and unclear project scoping lead contractors to bid higher
• Station design: Overly expansive stations with unnecessary rooms and nonstandard designs inflate costs

While prima facie one might see these as relatively separate problems, a careful reading of this report suggests they largely arise from one core cause, namely, the unchecked growth of the administrative state in both size (absolute number of employees and departments) as well as powers (arbitrary rule-making from petty bureaucrats not held to account by any central force). One can easily imagine an alternative to the status quo, where a smaller, more constrained, but also more efficient state, with levels of internal alignment comparable to those of a well-run multinational corporation accomplishes feats of productivity which far surpass the capacity of today’s New York City MTA.

Following the structure of the New York Case Study, we examine these independent reasons for cost overruns and show that they can be brought back to our central thesis of governmental bloat.

Inter-departmental strife within the government

The first reason presented by the authors of the New York Case Study is, put simply, a level of intergovernmental disarray that begins to resemble outright conflict more fit for a failed state than for one of the world’s greatest nexuses of culture and commerce.

One particularly egregious example is the usage of Marx Brothers Playground, a park on Second Avenue close to the sites of construction, to stage construction of the new 96th Street Station. Although this park is public land, the Metropolitan Transport Authority (MTA) paid a separate division of the New York City government, the New York City Department of Parks & Recreation (NYC Parks) $15 million for the right to use a section of Marx Brothers Playground for construction activities.

Such an arrangement is nothing more than a naked (albeit rational) attempt at value extraction between different limbs of the same organism, and a comparison to standard business practices makes clear the absurdity of this situation. The scale of such projects is hardly foreign to industry; for example, Shell Global reported combined expenditures of $35 billion in 2021, almost surely including large, complex, capital-intensive projects requiring deep interdepartmental cooperation. Yet even so, it is difficult to imagine, say, one division at Shell charging a separate division $15 million to use a piece of equipment it has in storage for the sole purpose of extracting as much of the overall project budget as it can. Were something like that attempted, the vice presidents responsible would likely not survive past the next quarter.

Another notable example is the MTA’s attempt to finalize plans for replacement of water utility lines in the construction path with NYC’s Department of Environmental Protection (DEP). Although the MTA had already contracted out the actual utility replacement job, the DEP refused to finalize and approve the MTA’s plans, largely because it insisted on replacement of 48-inch cast-iron pipes with larger, 60-inch steel pipes. Negotiations between intermediate managers of the two agencies stretched on for six full months, with 5-figure contractor costs accruing day after day, before the respective department heads eventually met in person and agreed to compromise on 48-inch ductile-iron pipes.

This is, of course, yet another example of one agency trying to extract part of the subway extension project’s budget for itself, albeit in a slightly less shameless fashion. Again, it is challenging to imagine an analogous situation happening in even the most dysfunctional of corporations. If a crucial component of a major project which is the primary and overriding focus of the company were delayed by anything approaching a full month because of some mundane interdepartmental dispute about pipe specifications, some higher-up executive, perhaps even the CEO, would almost certainly force the two department heads responsible to come to terms with each other. While disagreements certainly do happen, only in the most abysmal companies would mid-level managers argue with each other for half an entire year without either appropriate escalation or resolution of the problem at hand.

The case study supplies many other examples of interagency discord (e.g., refusal to disclose utilities maps on “grounds of national security” or insistence upon the inclusion of an excessive number of “back-of-house space” in new stations). Without needing to describe each one in explicit detail, the underlying theme is clear: rather than collectively aiming at a single goal, each of the separate parts of the NYC government appear to operate as independent fiefdoms largely concerned with optimizing their own finances and sometimes eschewing even the appearance of mutual cooperation.

One might initially be tempted to understand this as a problem of mere coordination rather than a problem with the scope of government itself. However, I believe such a perspective would be misguided. Scaling any organization while retaining a sense of focus on collective goals is challenging, a problem worsened by the relative lack of market competition between different forms of governance; the creation of a large administrative state is therefore intrinsically setting itself up for failure.

Indeed, one might even say that framing this as a coordination problem presupposes the existence of separate departments that need coordination in the first place! Division into focused departments in general is of course reasonable, but the divisions grow ever more fine as the diversity of tasks undertaken by the government increase and broaden in scope. The point here is subtle but important: imagine, for example, the nature of geographic divisions (states, counties, cities, and so on). One would surely agree that there are fairly ‘natural’ levels of separation here; countries separated by oceans or with vastly different cultures probably benefit from having separate governments, and while state boundaries in America are fairly arbitrary, they are ‘reasonable enough’ in an intuitive sense. With these geographic divisions, however, one much more easily sees the ridiculousness of overly fine subdivisions: if America were divided into five hundred or even five thousand states, for instance, the territorial fights, management of border crossings, and general logistical complexity seems prima facie much worse than a hierarchical model where smaller counties are grouped together into larger states.

At least theoretically, then, there is little reason why we would not expect the same problems to apply to non-geographic divisions of administrative labor within a given jurisdiction, as is the case with the separate departments of the New York City government. Being non-geographic in nature, however, and therefore somewhat more difficult to understand as a whole (one cannot exactly visually draw each department’s domain on a colored map), it is harder to determine whether or not city departments have over-proliferated at a single glance. “The proof is in the pudding,” thoughーit is hard to deny on the basis of actual, realized outcomes that this is in fact the case.

Crucially, as the overall scope of administrative labor increases, the level of subdivision also increases, whether explicitly or implicitly. To some point, of course, this increases efficiency; no reasonable person would argue that the entirety of Manhattan should be managed by a single, flat, structureless organization. Again, though, we are clearly well past the optimal point. Different departments of the same city government fighting to extract money from each other makes one imagine a poor man suffering from a stroke who cannot help but strangle himself with his right hand while his left attempts to resist.

One might well ask: doesn’t that mean we should just impose stronger top-down management or combine departments together under a single umbrella? In fact, separate departments all report to the head of the city and state governments, so isn’t this just a question of executive will and power? This is in fact how the authors of the case study frame the solution: “stronger leadership from a governor or mayor … could have avoided these delays and added costs.” To me, though, this has the feeling of inverted logic. Unfortunately, we cannot simply summon in great, charismatic leaders wherever we need them; if it were really just a matter of desiring, in the abstract, ‘better leadership,’ we would hardly have these problems to begin with! Instead, a more productive way to conceptualize these problems is to recognize that we are given the leaders that we have (whether they are good or bad), and that the degree to which they can exercise power is intrinsically limited by the overall scope of governmental activity.

In a sense, these deeply maladaptive outcomes can be argued to be the natural result of certain traits unique to the nature of administrative government. In a company, for example, while access to certain resources might fall under the management of a specific department, the core function of most departments (with the exception of human resources or compliance) is to accomplish some concrete task. In contrast, rule-making is an intrinsic part of the reason why many subdivisions of city government exist, and it is a primary means through which they exercise their power. The administrative rules that one department sets are then imposed, in a binding manner, on other departments, leading to a sort of ‘exponential blowup’ in the aggregate complexity of the regulatory thistle even if the nominal size or authority of government only grows roughly linearly. The existence of this impenetrable thicket of interlocking rules is then used to argue for more bureaucratic growth (to navigate the rules and, inevitably, create even more complex rulesets), but that is largely a distraction from the fundamental problem, which is that one department should not in general be able to completely stall the work of a separate department for months or even years due to the imposition of some capricious regulatory objection!

Overall, we propose that the effectiveness of the administrative state has reached a point of negative marginal returns on growth due to the intrinsic challenges of scaling, especially in the absence of market competition. When the administrative state grows too large, it inevitably fractures into a squabbling mess of overlapping subdivisions which, lacking any real sense of integration into a larger whole, naturally seek to expand their own fiefdoms rather than prioritize the provision of public goods and services. The most natural way to curtail these excesses is then to simply reduce the overall scope of government, which intrinsically increases the unifying power of the executive.

Excessive reliance on external contractors

Another source of gross inefficiency highlighted by the New York Case Study is an over-reliance on external contractors. In principle, the flexibility and modular nature of external contracting work could be an important cornerstone of an effective yet small government; however, we will see that the unique pathologies on display here again lead back to our thesis of administrative bloat.

The authors of the report supply a brief overview of the history of government contracting: starting in the 1960s, United States government agencies at all levels began to increasingly prefer contracting out projects to external agencies, which engage in a competitive bidding process, rather than retaining full-time, in-house staff. They refer to this, following the work of John J. Dilulio, as a “Leviathan by Proxy” model.

The advantages of relying on external contractors and consulting agencies are clear. They can be scaled up and down in a relatively flexible manner; they can respond to all kinds of specialized needs, including those for which it would be difficult to build out internal expertise; finally, consultants serve as a politically neutral “buffer” which lubricates interagency relationships and complex negotiation processes.

As the case study reports, the disadvantages are even clearer. The overhead paid to contracted labor is about 200% compared to the prevailing market wage for that expertise. This is high even compared to lavish government benefits, which typically result in an overhead of 100% (as compared to a more typical 30% in private industry). This is not unreasonable when the types of expertise required are so specialized that it does not make sense to hire them full-time; however, when consultants are tasked with “solving basic problems,” presumably using a level of expertise general enough to be applied across many projects in the agency’s scope, it is pure waste to continually rely on contracted labor rather than building out an in-house team. Comparing wages alone actually understates the inefficiency of this practice: the nature of government is to manage a geographic jurisdiction over an indefinitely long period of time, meaning that the value of context-specific, historical knowledge and unwritten expertise are particularly high. When “it [is] the consultants who [know] everything about past projects rather than agency staff,” the government is then forced into a vicious cycle of eternal reliance upon contractors as a de-facto overpaid, perpetually present branch of government.

Reliance on contracted work, which is typically structured around predefined contracts, also exacerbates failures elsewhere in the system. When different government agencies cannot come to rapid agreement on contentious issues, causing delays or revisions to planned work, these translate into expensive “change orders” to preexisting contracts. In the case of the dispute over water utilities pipe replacement previously described, the actual work of moving the utilities was contracted out, and even though the ultimate resolution reduced the scope of work to be done, the contract ended up 20% more expensive than the original budgeted cost. Even cost-cutting measures like simplified station redesigns can result in change orders that consume a substantial fraction of the money saved, such as the reduction in the 72nd Street Station’s width, which was estimated to reduce costs by $90 million but resulted in a $26.5 million change order.

At first glance, then, this might appear to be an argument for growing the size of government. After all, isn’t that what it means to hire and retain internal expertise?

Not necessarily so. One clue lies in the term “Leviathan by Proxy,” which implicitly reveals what we intuitively know to be true: even though these contracting agencies are not actual parts of government, they may as well be. There are only very few construction firms which have the requisite expertise combined with a tolerance for tedious government contracts, and some projects receive as few as one or two bids. In addition, the shift of historical knowledge from government employees to long-term contractors effectively ties them together at the hip; in fact, there is likely a “revolving door” where government workers who build up enough internal expertise can leave to work for an external agency, which then uses that knowledge to extract greater rents from the government. In effect, we have more or less the same people working on the same projects over long durations of time and with relationships no less contentious than actual inter-agency relations! Even if we nominally shift labor from inside the government to outside the government, isn’t that closer to a vacuous accounting trick than an actual reduction of governmental bloat? The proper response to accounting tricks is to simply ignore them, which leads us to straightforwardly conclude that hiring employees internally actually reduces the overall size of government while producing obvious gains in net productivity.

It is also important to not “miss the forest for the trees.” Why have we ended up in an equilibrium where government agencies at all levels seem to be addicted to farming out monolithic projects to inflexible contracting agencies? While government employees are hardly brilliant, those who stand atop the bureaucracies are not necessarily stupid or incompetent either, and we would do well to understand why they have chosen to operate as they do. Beyond the self-propagating nature of decaying institutional expertise, it is possible that regulatory barriers imposed on contract bid solicitation increase the fixed cost of setting up a contract to burdensome and arduous levels, thereby incentivizing an “all-or nothing” approach where an agency consumes either very little contract labor (full in-housing) or very large amounts (if you’re setting up a contract, may as well “make the most of it”).

The case study’s discussion of failings in the MTA’s bid procurement and risk allocation practices supports this viewpoint. For example, they describe state regulations on cost overruns as being counterproductively heavy-handed:

Fourth, state regulations are heavy-handed whenever costs run over prior estimates. The Cuomo administration added a debarment rule blacklisting contractors whose final costs go 10% above their bids. In a letter to the MTA Board, the Citizens Budget Commission (2019) argued that the threat of debarment reduced competition because contractors were afraid to submit bids on MTA projects and risk being banned from bidding on future New York State contracts. One source at the MTA compared the threat of debarment as the state pointing a gun at contractors, for which they respond with a bazooka, bidding anything from 15% to 40% higher, the latter including all risks (Personal Interview O 2021).

One immediate observation here is that excessive and arbitrary regulation, disconnected from on-the-ground reality, leads to the imposition of binding rules that have the opposite of the intended effect! As the study notes, the rule in question originates from the Cuomo administration itself rather than from a separate department; here we see that a stronger executive may not in and of itself resolve more problems than it solves. A more sustainable solution is to aim for true unification and goal alignment through reduction of the overall size of government.

More generally, though, there exist a variety of practices, both explicit and informal, that burden contractors and lead to overbidding, such as “[putting] the onus of underground geotechnical risk on … contractors” or the withholding of the MTA’s own cost estimates from bidders rather than letting private bidders reference the contracting agency’s internal estimates. While these are typically framed as impositions upon the contractor rather than the contracting agency which then causally lead to expensive bids, it is crucial to realize that even during the initial bidding process, the burden is shared in equilibrium: these frictions cause delays, headaches, and complications for both parties in the process.

As a result of the accumulation of innumerable frictions in the contracting process, we then tend toward a state of affairs where engaging contract work at all means that you may as well “go all the way.” This is easy to see by imagining what an effective usage of contractors would look like: tightly scoped projects, with key consultants brought in at the crucial moments to supply necessary expertise. However, if the negotiation and contract initiation process is so fraught with inefficiencies that it takes months to establish a simple relationship, this is simply not possible! The outcome then tends toward expansive contracts which seek to anticipate, often imprecisely, future needs and bundle them together with more basic aspects of the planning or construction process which could have otherwise been effectively in-housed. This then leads to change orders down the line, which then beget even further rule-making about cost overruns, and so on…

One might reasonably ask: could things be worse off on net without the large regulatory framework we have established around external contracting? After all, the rules are there for some reason. It is certainly imaginable that a loosening of the rules would lead to a greater amount of unethical kickbacks and corruption-adjacent activities where separate parties cooperate together to milk money from the largesse of the state. That may very well be the case! Here, history offers a useful guide.

Interestingly, in both America’s own Gilded Age as well as in post-Deng Chinese industrial development, there appears to be an optimal, nonzero amount of corruption that “greases the wheels” by aligning incentives toward the end goal: on-time completion of revenue-generating projects. Notice that in the phrase “separate parties cooperate together to milk money from the largesse of the state” we have the essential component of separate parties cooperating togetherーarguably not a state that prevails in modern-day infrastructure construction! As Marginal Revolution comments:

1. Access money dominates.

More concretely, politicians prosper by getting things built, not by preventing things from getting built.

2. China’s political system operates on a profit-sharing model.

3. Capacity-building reforms have curtailed damaging forms of corruption.

4. Regional competition checks predatory corruption, spurs on developmental efforts, and ratchets up deals.

The Cuomo administration’s rule on cost overruns may therefore be seen as part of a larger phenomenon: that of well-intentioned rule-making which attempts to rein in what superficially appear to be wasteful excesses but in fact end up counterproductively introducing deeply harmful frictions that clog the entire state apparatus and sabotage its ability to contract out work in an efficient manner.

Closing thoughts

Broadly speaking, the ability of man to execute at scale and to rationally shape complex, human-involved systems is quite low. Therefore, we should recognize these weaknesses in a forthright way and seek to avoid hubris by preferring smaller, more constrained governmental scopes.

In the end, it is difficult to know with certainty whether these problems would have been alleviated with a more tightly run, constrained New York City government. However, as one reads the New York Case Study, it is difficult to avoid drawing mental comparisons to the level of efficiency which prevails in a typical corporation and to ask: why are things the way they are? Inevitably, as we trace the threads of these disparate problems, we find that they seem to originate from a common source: a remarkable degree of self-inflicted bloat.