The degree of concentration of economic activity in urban areas is striking, as they host more than 50% of the world’s population on only an approximate 2.7% of the world’s land. There is a consensus among planners and policymakers, however, that even higher densities within cities and urban areas are desirable, at least, on average (OECD 2012). Most countries pursue policies that implicitly or explicitly aim at promoting ‘compact urban form’, reflecting the concern that unregulated economic markets will fail to deliver efficient and equitable allocations of uses and infrastructure.
From a policy evaluation perspective, the critical question is whether the dominating ‘compact city’ policy paradigm, which aims to shape the habitat of the urban population over the decades to come, can be substantiated by evidence. This question is notoriously difficult to answer since density effects presumably materialize in a broad range of outcomes, such as accessibility (to jobs and amenities), productivity, innovation, rent, various environmental outcomes (open space preservation, biodiversity, pollution reduction, energy efficiency), efficiency of public service delivery, health, safety, social equity, transport (ease of traffic flow, sustainable mode choice), and self-reported well-being.
While urban economics research has made great progress in understanding the effects of density on productivity and wages (see Combes and Gobillon 2015 for a summary), the literature on most other outcomes is less developed. Moreover, the evidence is scattered across various separate literatures. To date, there is no accessible summary of positive and negative density effects on a broad range of outcomes.
In a recent paper, we aim to fill this gap by providing a synthesis on the state of knowledge on the economic effects of density, summarising the existing evidence, and complementing it where needed (Ahlfeldt and Pietrostefani 2019).
Building an evidence base
We begin by reviewing the existing literature and compile a unique evidence base that contains 347 estimates (from 180 studies) of density elasticities of various outcomes. Along with the density elasticities, we encode various study characteristics such as the publication date and venue, the geographic origin, and layer of analysis, among others. To capture the rigour of the analysis, we encode the method on a Scientific Methods Scale (WWC 2016) that ranges from ‘0: purely descriptive’ to ‘4: exploiting plausibly exogenous variation’. As a measure of impact, we construct a year-since-publication-adjusted citation index using data from Scopus and Google Scholar.
Next, we enrich this evidence with original elasticity estimates where the evidence base is thin or inconsistent. We provide transparent density elasticity estimates based on a consistent econometric framework and OECD data that refer to 16 distinct outcome variables. For some outcomes, such as the density elasticity of preserved green space, our estimates are without precedent.
Recommended density elasticities
We then condense the evidence reviewed as well as our original estimates into a set of recommended density elasticities for a range of outcome categories listed in Table 1. Specific to each category, we either recommend a citation-weighted mean across the elasticities in our evidence base, an estimate from a high-quality original research paper, or one of our original estimates. We believe that these recommended elasticities represent a potentially useful resource for researchers wishing to explore the welfare effects of policies related to economic density, be it in back-of-the-envelope calculations or calibration of structural models.
There are limitations, however. The quality and quantity of the evidence base is highly heterogeneous across categories. We strongly advise to consult section 4 in the appendix to our paper (Ahlfeldt and Pietrostefani 2019), which provides a discussion of the origin of each of the recommended elasticities against the quality and quantity of the evidence base, before applying any of the elasticities reported in Table 1 in further research.
In a nutshell, we see sufficient evidence that seriously engages with separating the effects of density from the effects of correlated unobserved fundamentals to allow for a causal interpretation in the following categories: 1 - Wage and productivity, 2 - Innovation, 3 - Rent, 4 - Vehicle miles travelled, 10 - Pollution reduction, and 12 - Average speed.
For the other categories, the estimated elasticities are better interpreted as associations in the data. Any causal interpretation will rest on the strong assumption that density has been determined by historic factors that no longer have effects on contemporary outcomes. We stress that significant uncertainty surrounds the effects of density on income inequality, urban green, health, and self-reported well-being. In general, the recommended elasticities are best understood as describing area-based effects that include composition effects.
Table 1 Recommended elasticities by outcome category
Notes: Based on Table 6 in Ahlfeldt and Pietrostefani (2019). Density elasticities are best understood as referring to large cities in high-income countries. In general, they represent correlations and not necessarily causal estimates. See section 4 in the appendix to Ahlfeldt and Pietrostefani (2019) for a critical discussion of the evidence base by category.
We acknowledge that for most outcomes the density elasticity is likely context-dependent. While, given the quality and the quantity of the evidence in the literature, we constrain ourselves to recommending point estimates in Table 1, our analysis also reveals insights into some interesting dimensions of heterogeneity.
Studies that are more frequently cited, or use more rigorous methods, find less positive density effects (in a normative sense). The estimates also become less positive over time, possibly reflecting a trend towards the application of more rigorous methods. This confirms that identifying high-quality evidence is crucial for evidence-based policy making.
Our analysis also reveals some insights into geographic heterogeneity in density elasticities. For non-high-income countries, the estimated density elasticity of wages, at 0.08, is twice as large for high-income countries, on average. Mode choice is less likely to change with density for non-high-income countries, whereas the gains from density in terms of domestic energy consumption appear to be larger. Compared to other developed countries, density in the US is associated with larger skill wage gaps and higher rather than lower crime rates. Our review of the literature also suggests that the effect of density on rents may not be log-linear. The density elasticity of rent increases by 0.063 for every increase in population density by 1,000 inhabitants per square kilometre. We do not find a similar non-linearity in the effect of density on wages, suggesting that it is an agglomeration cost that increases at an increasing rate that limits the growth of cities.
Takeaways for policy
In the last step of our analysis, we monetise the economic effects of density to provide some more explicit policy guidance. We compute the per capita present value (PV, at a 5% discount rate) of the effect of a 1% increase in density for a scenario that roughly corresponds to an average metropolitan area in a developed country (Table 2). For this purpose, we combine our recommended density elasticities from Table 1 with several valuations of non-marketed goods such as time, crime, and mortality risk, or pollution, among many others.
Table 2 Present valuea effects of a 1% increase in density (US dollar)
Notes: Based on Table 8 in Ahlfeldt and Pietrostefani (2019). a The present value per capita for an infinite horizon and a 5% discount rate. All values in $. b Amenity equivalent of after-tax wage increase assuming a marginal tax rate of 32% as in Albouny and Lue (2015). c After-tax wage increase as discussed in b. d Excludes $19.18 of dirving energy cost ($0.15/mile gasoline cost) discounted at 5%, which are itemised in 11. e Assumes a 10.2% elasticity to avoid double-counting of road trips already included in 4. f Amenity effect, excludes health effect itemised in 14. g Set to zero to avoid double-counting with 11. We assume that public services are nationally funded.
An important insight from Table 2 is that the effect on rent exceeds the effect on wages. In a spatial equilibrium framework (Rosen-Roback), this implies that individuals are willing to give up on real wage in order to live in denser cities, so density must be a net-amenity. There is also a positive net external (to the city) welfare effect, which is primarily driven by the lower cost of providing (nationally funded) local public services (column 6). However, summing up the monetary equivalents of all amenity and dis-amenity categories (sum in column 3) results in a positive value that is smaller than the ‘compensating differential’ (sum in column 2). While density seems to be a net amenity, Table 2 suggests that part of the rent increase may be attributable to the higher cost of providing space in addition to enjoyable amenities.
Coming back to policy implications, our results indeed substantiate that policy-induced densification may lead to aggregate welfare gains. However, there may be a collateral net-cost to renters and first-time buyers if residents are not perfectly mobile and housing supply is inelastic.
References
Ahlfeldt, GM, and E Pietrostefani (2019). "The economic effects of density: A synthesis", CEPR Discussion Paper 13440.
Albouy, D, and B Lue (2015). "Driving to opportunity: Local rents, wages, commuting, and sub-metropolitan quality of life", Journal of Urban Economics 89: 74–92.
Combes, PP, and L Gobillon (2015). "The empirics of agglomeration economies", Handbook of Regional and Urban Economics 5: 247–348.
OECD (2012), "Compact city policies: A comparative assessment", OECD Green Growth Studies, OECD Publishing.
What Works Centre for Local Economic Growth (WWC) (2016), "Guide to scoring evidence using the Maryland Scientific Methods Scale".
