Affordable housing law and policy in an era of big data.

Author:Davidson, Nestor M.
Position:Special Issue on Affordable Housing

Introduction 277 I. Why Might Big Data Matter to Affordable Housing Policy? 281 A. Big Data and Policy 281 B. From Outputs to Outcomes in Affordable Housing 284 II. Big Data in Affordable Housing Law and Policy 287 A. Some Examples of Big Data's Potential in Affordable Housing 288 1. Siting Decisions, Mobility, and Neighborhood Effects 288 2. Housing Portfolio Management 290 a. Subsidy Targeting and Market Conditions 290 b. Enforcement and Housing Quality 291 c. Unifying the Subsidized and Unsubsidized Housing Portfolio 293 3. Resident Relations and Services 293 B. Synthesis and Reflections on the Role of Law 295 III. Big Data's Dark Side: Caveats and (Some) Responses 297 Conclusion 300 INTRODUCTION

Every year, federal, state, and local governments invest more than $50 billion to provide housing for people who cannot otherwise afford shelter. (1) In addition to this housing assistance, policymakers also make a variety of choices that impact the landscape of affordable housing, including in zoning, infrastructure, housing finance more broadly, and in a myriad of other policy domains. (2) These policies can make a profound difference for the millions of individuals and families helped, but are too often undertaken with only the vaguest, visceral sense of their consequences beyond the bare facts of putting roofs over people's heads.

However, affordable housing policy is beginning to experience a shift in perspective. To the extent that policymakers have collected data on impact, the focus traditionally has been primarily on outputs. These measures included the number of units built through a given investment, the number of people served under a given program, the number of construction or property management jobs created, and the like. But outputs are not always--indeed, not often--the same as outcomes, the actual short- and long-term consequences of policy interventions for those served by affordable housing programs and the communities at issue. (3)

In recent years, researchers and policymakers have begun to evaluate the results of policy interventions for people in subsidized housing on measures such as income, educational achievement, physical and mental health, and even subjective wellbeing. (4) Rather than merely track whether people have housing at a given level of affordability, this new focus understands that housing is a platform for a variety of life outcomes and that housing policy choices can meaningfully impact the arc of those outcomes.

This emphasis on outcome measures reflects a broader embrace of the use of data for decision making by managers and policymakers across the private and public sectors. (5) The ability to collect data in a more rigorous and systematic way and the development of tools to make that information actionable--particularly to make transparent patterns that would otherwise be opaque--is beginning to change a range of decisional processes. This shift can been seen in everything from how professional baseball teams select players to how Facebook, Google, and other companies target their advertising to how investors seek value. (6) And data increasingly means "big data," an admittedly fuzzy (and arguably hackneyed) concept that roughly refers to the use of relatively large data sets, often aggregated across previously disconnected areas, mined for the predictive value of underlying patterns and trends. (7)

Although much has been written about data-driven policymaking, the specific role of the legal system in improving program design and implementation deserves deeper exploration. (8) Across a range of affordable housing examples, explored below, (9) new data tools are starting to sharpen policy, but also creating a positive feedback loop in which agencies provide data to grantees to shape how they implement policy, gather more information about outcomes, and then share all of that information with other regulators or advocates to help advance other legal mandates, notably around enforcement and private rights of action. The confluence then between emerging analytic tools and a deepening understanding of the connection between inputs and outcomes makes affordable housing a particularly fruitful policy arena in which to explore law's potential to generate as well as facilitate the deployment of data to improve policy. (10)

New data analytic tools are by no means a panacea. Incorporating big data into affordable housing law and policy raises serious concerns that are well rehearsed in the literature but worth reiterating in this context. On a structural level, data quality and integrity poses a basic challenge, as does the perennial risk that policies driven by what can be measured, such as bricks-and-mortar metrics, will inherently misdirect resources away from goals that are less quantifiable, such as policies that reflect the felt experience of people living in communities of concentrated poverty. Legal scholars have also raised important concerns about privacy and data security. And perhaps most importantly, there is a genuine risk that the very people whose lives should be at the center of affordable housing policy will be even more marginalized than they currently are when measurability is privileged over meaning.

There are potential answers to each of these concerns. Data must be taken in context and with appropriate skepticism; data should be anonymized and aggregated to the extent possible; and policymakers must find ways to be sensitive to the dignitary harms that attend quantification. But an alternative that fails to measure impact is hardly palatable and the potential benefits of enhancing the ability of policymakers to understand the consequences of their actions can actually advance the dignitary interests of those served by affordable housing policy.

This all may seem somewhat technical and dry--rarely has the heart fluttered for the ephemeral value of, well, analytics. Those motivated to engage in affordable housing--or other areas of poverty law and policy--by compassion may recoil at discussions of number crunching. That is entirely understandable, but data and compassion need not stand in opposition. For policy to be effective, we need both.

The Article is organized as follows. Part I describes big data's potential for policy making and the emerging shift in affordable housing policy from outputs to understanding broader measures of outcomes. Part II then looks at several areas where data analytics are either currently changing approaches to affordable housing policy or where there is particularly strong potential for such a shift--by no means an exhaustive list, but illustrative. From these examples, the Part then synthesizes what this reveals about the role of law in generating data and deploying data. Finally, Part III highlights reasons for caution and some responses.

Before turning to the substance of this Article, I want to break the authorial fourth wall for a moment, to address a question about our current context. This Article grew in part out of my experience working in the U.S. Department of Housing and Urban Development ("HUD") General Counsel's office early in the first Obama Administration. (11) There, I worked with the team that developed the agency's affirmatively furthering fair housing ("AFFH") rules. I was also exposed to the agency's work, discussed below, to embrace data-driven decision-making more generally. (12) Since beginning this Article, however, the election of Donald Trump has injected uncertainty into affordable and fair housing policy, especially with Dr. Ben Carson's appointment as Secretary of Housing and Urban Development. (13) To the extent that Dr. Carson has a record with respect to HUD's mission, it consists primarily of voicing skepticism about the ability of the government to address segregation, (14) although Dr. Carson in his confirmation hearing seemed more open to the work of the agency. (15) This, together with the Trump Administration's general skepticism about regulation, might cast some of what this Article argues in doubt, although it is certainly too early to know. But it is not clear that data will be less important even if HUD moves away from promoting integration, economic opportunity, deconcentration of poverty, and other goals broader than simply subsidizing housing. Moreover, the longer-term trend will still favor data-driven decision making, state and local efforts will continue, and seeds planted in recent years may flower in other domains.


    This section provides a brief primer on big data's potential for policy improvement. It then turns to the start of the embrace of outcome-driven policymaking in affordable housing. Big data and outcome-driven policymaking, of course, are not synonymous--the former has many applications and the latter does not require novel data tools--but their intersection raises intriguing possibilities.

    1. Big Data and Policy

      To state the obvious, policymakers have long relied on data to make decisions. Indeed, categorization, record-keeping, sorting, and similar bureaucratic informational management tools have so long been inherent to the administrative state that we rarely pause to note the ubiquity of the phenomenon. (16) In this sense, modern policy has always been data-driven, (17) and a broad array of policymakers have embraced the idea of evidence-based interventions. (18)

      What is changing today is the availability of significantly more data and more powerful computing capabilities. (19) Big data is one of those cliched terms that is hard to pin down precisely, and it is not necessary to do so here. Generally, though, the term encompasses a set of related phenomena. (20) First, big data refers to the capacity to collect and aggregate massive datasets and similar information--this is what is "big" about the phenomenon, as opposed to, well, just data. (21) Before I started research for this Article, I...

To continue reading