How Costly Is Permitting in Housing Development?

How Costly Is Permitting in Housing Development?

Abstract

Permitting costs are widely cited, but little analyzed, as a key burden on housing development in leading U.S. cities. We measure them using an implicit market for "ready-to-issue" permits in Los Angeles, where landowners can prepay permitting costs and sell preapproved land to developers at a premium. Using a repeat-listing difference-in-differences estimator, we find developers pay fifty percent more (forty-eight dollars per square foot) for preapproved land. Comparing similar proposed developments, preapproval raises the probability of completing construction within four years of site acquisition by ten percentage points (thirty percent). Permitting can explain one third of the gap in Los Angeles between home prices and construction costs.

One Introduction

The cost of housing is one of the top economic issues facing large, highly productive U.S. cities. In places like Boston and Los Angeles, rents and home prices are substantially higher than the national average. Higher costs of living offset these cities' higher nominal incomes, leaving their residents not consistently better off than in lower-income, lower-cost cities. Moreover, scholars have estimated that the lack of housing in such cities meaningfully reduces aggregate U.S. output and most greatly burdens the urban poor and middle class.

Among all large U.S. cities with current monthly rents above two thousand dollars on a typical apartment, none issued permits for more than five new homes per one hundred residents cumulatively from twenty fifteen to twenty twenty-five. Why haven't these cities built their way out of such high rents? An academic literature has argued for the importance of restrictive land-use regulations, rather than high construction costs, in limiting a supply response. The most prominent evidence for this view is a series of articles, starting with Glaeser and Gyourko, that finds large gaps between house prices and estimates of their construction costs, consistent with binding regulatory constraints.

A challenge facing both scholars and policy analysts is that the regulatory environment in real estate development is complex, involving not only explicit taxes, limits, and mandates but also a set of diffuse procedural burdens. One of the most widely-mentioned sources of burden is permitting, or obtaining legal approval to build. Policy analysts have argued that permitting exposes projects to significant risk and delay. Such arguments have fueled debate about an "abundance agenda" and a wave of reforms intended to simplify permitting. Yet it remains unclear whether, in high-cost U.S. cities, permitting is a minor nuisance or a major impediment to housing supply. In particular, many of these cities also have restrictive zoning rules and high construction costs, forces that may be a far larger burden than permitting on its own.

Measuring permitting costs is challenging for several reasons. First, the costs are unobserved and likely have both time and resource components. Developers' compliance expenditures are proprietary information, nor is it clear what cost of capital should be used to value permitting-related delays and risks of project failure. Second, permitting costs are equilibrium objects, not policy parameters. Developers propose structures with permitting costs in mind, creating a selection problem that complicates the interpretation of permit data. Third, and most fundamentally, rights to build are tied to specific properties, such that permits are never traded separately from land and structures. Unlike in "cap-and-trade" pollution markets, therefore, economists lack an explicit permit price that would reveal compliance costs of marginal projects.

This paper measures permitting costs by studying an implicit market for development approval. In Los Angeles County, landowners may secure all necessary permits for a proposed project before sale, transferring the bundle of land and "ready-to-issue" (RTI) permits to developers. In theory, the premium paid for this bundle over raw, unpermitted land reveals developers' marginal willingness to pay to skip the permitting process. The RTI market is economically significant and mature: In some neighborhoods, it accounts for one in four land sales and one in ten properties suited for redevelopment ("likely teardowns"), with three hundred fifty-three million dollars in transaction volume in twenty twenty-four. Exploiting this market, we estimate the approval premium, the impacts of preapproval on time-to-build and project completion, and permitting's share of the total regulatory tax on new housing.

To study the RTI market, we assemble a dataset of properties for sale advertised on the Multiple Listing Service (MLS) for Los Angeles County from nineteen ninety-five to twenty twenty-four. Our data contain ninety-five thousand seven hundred twenty-four unique land and likely-teardown parcels, of which we observe five point three percent (five thousand ninety-two) are listed with and without permits. We use large language models (LLMs) to classify permit status from the text descriptions that real estate agents provide to the MLS, and we validate the LLM output with thorough human review. Finally, we link listings to permits, property tax assessments, and zoning rules, allowing us to estimate time-to-build effects and perform supplementary analyses.

We begin by characterizing the permitting process in Los Angeles and the city's land market. Relative to other U.S. cities, permitting in Los Angeles appears burdensome: We estimate that building a midsize apartment complex takes around twice as long there as in Raleigh, North Carolina, or Fort Worth, Texas. On average in Los Angeles, time-to-permit accounts for forty percent of total time-to-build. Analyzing preapproval, we find it is more common on smaller lots and in denser neighborhoods, and it appears driven by specialized investors who acquire raw land specifically to permit and resell. While these facts lend support to our interpretation the approval premium as an equilibrium price of permits, they also raise the possibility of non-random selection into preapproval.

Our main empirical strategy is a repeat-listing difference-in-differences design. Among properties that are repeatedly listed, we compare price changes that coincide with changes in permit status to price changes on properties that remain raw land. By removing time-invariant characteristics, this design addresses a common challenge in cross-sectional hedonic regressions: unobserved fixed attributes that affect both prices and selection into preapproval. The capitalization of anticipated approval poses a second challenge. If properties pay permitting costs before approval, land prices should rise in advance,

potentially contaminating the repeat-listing design. We adopt an event-study design to account for this anticipation effect.

We find a substantial approval premium. In the repeat-listing design, permit approval raises the price of vacant land by fifty percent on average. The cross-sectional design yields a similar estimate. In dollar terms, the premium amounts to forty-eight dollars per square foot of land for the average property, or approximately thirty-six percent of construction cost. For parcels with pre-existing structures, approval premia are smaller than those on vacant land in percentage terms but comparable in dollars, particularly if we reweight the data to be more similar on observables. Sensitivity analyses show our estimates are robust to time-varying controls for neighborhood characteristics and other listing information. Finally, we show permit arrival is partly anticipated, with a rise in listing mentions of pending permits two years before approval, coincident with the time path of land-price capitalization. However, anticipation bias in our approval premia appears negligible, as most land sales occur several years before permit news arrives.

Linking listings to permits, we also study the effects of preapproval on remaining time-to-build using a semiparametric hazard model. As we do not have a repeat-listing design for this outcome, we rely on comparisons of observably-similar projects with and without approved permits at the time of site acquisition. We find that preapproved projects are eight to twelve percentage points more likely to be completed within four years of site acquisition, relative to a counterfactual four-year completion rate of thirty-five percent. To probe selection bias, we vary the set of control variables. Sensitivity analyses suggest that preapproval robustly increases the probability of completing quickly (i.e., within three years), whereas the impacts on ever completing are more sensitive to controls. Benchmarking our estimates to the average time-to-permit of preapproved properties, we find preapproval is effectively a one-to-one transfer of waiting time from developers to landowners.

To interpret these results, we propose a simple equilibrium model of housing development, wherein developers choose how much housing to supply and how much effort to invest in permitting speed. We obtain three results from the model. First, the approval premium can be expressed as the sum of two objects we call "pure wait" and "capitalized hassle," reflecting respectively the time and resource costs of permitting. Second, the effect of preapproval on remaining time-to-build allows us decompose the approval premium into these objects. Third, we relate permitting costs to the overall housing cost wedge, allowing us to contextualize permitting's importance within land-use regulation more broadly. The third result shows the permit share of the wedge can be measured either exclusively with land-price data or by the Glaeser and Gyourko house-price approach, which allows us to relate the approval premium to the city's overall regulatory burden.

Leveraging the model, we reach two conclusions about permitting in Los Angeles. First, at standard discount rates, pure wait accounts for much of the approval premium, suggesting that time-based measures of permitting can provide useful insight into its cost.

Second, on citywide average, permitting explains around one third of the gap between housing prices and construction costs, making it a key regulatory cost on development in Los Angeles. For such analyses, we exploit the size and richness of our listings data to compute detailed estimates of housing-price premia over construction costs, adapting methods from Glaeser and Gyourko and Gyourko and Saiz. Overall, our analyses reveal rich and important impacts of permitting on housing and land markets.

In doing so, this paper contributes to literatures on regulation in public finance, urban economics, and real estate economics. An outstanding issue in this area has been to connect two distinct empirical approaches to land-use regulation. The first, beginning with Glaeser and Gyourko, computes regulatory taxes as "top-down" wedges between housing prices and physical construction costs. While this approach documents the presence of such wedges across diverse urban contexts, it typically treats the internal composition of the wedge as a "black box."

A second tradition has sought to open this black box by compiling indices of land-use regulation across jurisdictions. These indices typically measure de-jure rules and questionably capture the de-facto reality of bureaucratic hassle. Also in this "bottom-up" tradition are studies of land-market capitalization of specific land-use regulations. In related work, researchers have also examined the effects of regulations on development. Perhaps most closely related to our paper is Diamond et al., which infers regulatory burdens from developer behavior within a structural model. Our Los Angeles ready-to-issue setting provides complementary market-based evidence about de-facto compliance costs that are plausibly important but challenging to measure.

Methodologically, our analysis builds on a tradition of hedonic methods for valuing non-marketed attributes. Our repeat-listing design fits within a branch of that literature which has used parcel-level panels, mostly for the valuation of environmental disamenities. Our study also exhibits two applications of machine learning and LLMs to hedonic estimation: isolating attributes that existing approaches (e.g., keywords) capture with considerable noise, and soaking up once "unobservable" potential confounds.

Finally, we relate to recent research on permitting processes for real estate and infrastructure. Manville et al. and Kestelman study permitting reform in Los Angeles, measuring its net impacts rather than total permitting costs. Our time-to-build analysis builds on efforts to measure the speed of approval and development and connects to studies on the causes of rising infrastructure costs. Where appropriate, we discuss how the richness of our data and context can inform the measurement of permitting costs in other real estate markets.

The paper proceeds as follows. Section Two describes our data and setting in Los Angeles. Section Three provides a model of permitting as a foundation for the empirical analysis. Sections Four and Five respectively estimate the approval premium and the effect of preapproval on remaining time-to-build. Section Six relates our findings to the overall housing cost wedge. Section Seven concludes.