The Biden administration will, it seems, have gotten or be about to get its coveted “bipartisan” infrastructure deal by the time this forum is published, by agreeing to divide into two parts the $2.65 trillion American Jobs Plan they laid out in some detail on March 31. They have gotten a first tranche past the filibuster by agreeing to limit their focus for now just to what those Republicans who signed on are calling “actual infrastructure,” defined as roads, bridges, highways, airports, broadband, and the like. Then, if all goes well (which it may well not), they will try to get at least some of what remains by way of “reconciliation,” that Senate quirk which allows for one spending bill a year to become law with the support of just a simple majority. This remainder, if we get it, is expected to include significant new spending to shore up what the administration is selling as the nation’s “care” and “social” infrastructures, to help “revitalize manufacturing, secure U.S. supply chains, invest in R&D, and train Americans for the jobs of the future”; or, as a July 29 Washington Post explainer would have it, it will be spending on “Democratic priorities not traditionally considered part of core U.S. infrastructure.”

Breznitz’s essay bears on this debate. He usefully debunks the misbegotten notion that every region can reasonably hope to—or even should want to—become the next “silicon somewhere,” and is right that even Silicon Valley firms are innovating in the context of a highly fragmented global production structure rife with new “opportunities for innovation-based growth beyond early-stage invention of a new product.” So, for instance, Silicon Valley firms that were at one time dominant in semiconductor fabrication have continued to prosper even as many have moved on in their technological focus, in part because they are specialized in “stage one” innovation—but also because they have strong relationships with suppliers in Taiwan that have taken on responsibility for the incremental improvements constitutive of “stage three” innovation. Those Taiwanese firms, meanwhile, have established themselves as global players and overseen a near wholesale transfer of a key segment of the industry from West to East precisely by trying less to compete with than to complement U.S. comparative advantage. They recognized—to borrow a phrase that circulates in the world of U.S. professional basketball—that it is a mistake to aspire always to be the star of one’s team; it is often better for all involved to try instead just to “star in your role.”

The other stories in the essay help as well to flesh out the argument and, although it is hyperbolic to claim “there are more paths open for innovation-based local prosperity than ever before” (ever is a very long time), it is clear that communities pursuing innovation-based inclusive prosperity in fact have “many more choices than the Silicon Valley evangelists would lead us to believe.” My complaint, to the extent I have one, is that Breznitz’s arguments end where the hard part begins. We read in the conclusion that innovation policy for local prosperity only works well if the “agents of innovation . . . are embedded in an economic ecosystem” that includes not only an education system that provides them with specialized skills, but also that helps to engender and institutionalize “active networks” of stakeholders from both industry and government. By implication, communities must “ensure the supply and creation of such collective goods” if they are to establish themselves as stars in their chosen role, fostering some capacities but leaving others to others. We are not told, however, either how those communities are to do this, or how it differs (beyond just its focus on stage one innovation) from what techbro purveyors of Silicon Somewhere strategies are selling instead.

I am familiar enough with Breznitz’s research to know that he has far more to say than he did. Still, given that we are in the midst of a very consequential debate in the United States over when and how to direct public spending in pursuit of the kinds of “connectedness” he says we need, I want to push him in his response to say some of it. It is not enough to observe that we have more choices than we sometimes think if we are unable to muster the political will to make those choices. It is obvious, of course, that the willingness of a supermajority of senators to sign on to spending on “actual” infrastructure is mostly driven by their ability to read polls. But, to the extent we are in the business here of trying to help people make good policy, I worry that we have better tools to justify government spending to help us move ourselves, our things, and our data from place to place than we do to explain why we also need spending on the social infrastructure that—if Breznitz is right—is no less critical to the economic well-being of communities hoping to take advantage of the opportunities Breznitz identifies.

It is easiest to get consensus on repairs to the United States’ tangible and visibly crumbling “actual” infrastructure of cracked bridges, flooding tunnels and the like, since these are subject to externalities and non-excludabilities and, as a result, to well-established “market failures.” The solution, when we agree that there is in fact a market failure, is to identify some combination of direct government provision of goods and services, or some mix of taxes and/or subsidies calibrated to “get the prices right,” to ensure that the decentralized decisions of people lead them to more or less do what we hope they will. So, for instance, we want most of those who commute from New Jersey to Manhattan to be able to take trains even after the next superstorm since they don’t actually all even fit on the roads; and we want Americans in rural areas to have access to, and to be able to afford, the broadband access they’ll need to work and study from home in the next pandemic.

When it comes to thinking through the repair of (in Biden’s terms) our “care” and “social” infrastructures, there are certainly some well-identified market failures around the direct funding of basic research and to spending on education (i.e., investments in “human capital”). But, if Breznitz is right, innovation-led growth requires more than that. It requires investments in the alphabet soup of programs tucked into the interstices of U.S. federalism oriented toward what Andrew Schrank and I have called “network failures,” where the aim is less to get the prices than the relationships right. These include programs such as ARPA-E for Energy, ARPA-C for climate (if, in fact, it gets off the ground), and the Manufacturing Extension Partnership, where “street-level bureaucrats” are given the discretion they need to blur boundaries between public and private, to help generate and sustain what Breznitz refers to as “collaborative public spaces—places where an industry moves from sharing knowledge to becoming a community.”