Housing is at the core of economics. It represents most households’ largest asset, and rent similarly accounts for a significant portion of inflation indexes as a major household expenditure. This column uses data from the Historical Housing Prices Project to produce a nationwide index for US house prices spanning the period 1890 to 2006. The data suggest that both owned and rental housing prices increased more over the 20th century than previously thought, and also that overall inflation over the period may have been underestimated.
Housing is a central issue in many economies around the world, with significant social and political implications. It is at the core of economics, too, with shelter the single largest component of household spending in many countries, and real estate the largest item on household balance sheets. As a result, questions about housing – including its affordability (Biljanovska et al. 2024), pricing (Schularick et al. 2015), and availability (Albuquerque et al. 2020) – are active topics of research.
Because of its durable and immobile nature, the housing system evolves slowly, and outcomes today reflect decisions and actions made over previous decades. To understand the current housing system, we must therefore take a long-run perspective. An emerging literature investigates housing market outcomes over the long run. This work includes diligent data-gathering efforts for specific markets, such as Korevaar (2023) on Amsterdam and Nicholas and Scherbina (2013) on Manhattan, and urban rents more broadly across a range of important global cities (Eichholtz et al. 2023). This literature also includes cross-country research that seeks to better embed the housing system, including mortgage financing, into an analysis of macroeconomic outcomes over time (e.g. Jordà et al. 2015, 2016) and related work on long-run housing prices (Knoll et al. 2017) and the rate of return on assets more broadly (Jordà et al. 2019).
Despite the centrality of housing to the American economy, long-run housing price series are limited, particularly prior to the 1970s for sales and for any period of the 20th century for market rents. Scholars interested in rental prices have used the Bureau of Labor Statistics (BLS) Rent of Primary Residence (RoPR) series, which is based on contract rents and used to construct consumer price indices. Scholars interested in sales prices over the long run have primarily relied on the pioneering work of Shiller (2015), who spliced together several data sources to obtain a national series beginning in 1890. At the city-level, annual housing price series that span the 20th century do not exist in the US, for either market rents or sales. Thus, we know relatively little about how the price of housing has evolved within and across US cities over the long run, which is fundamental to our understanding of housing markets and the US economy more broadly.
A newly released database addresses this gap. The new Historical Housing Prices Project, hosted by the Federal Reserve Bank of Philadelphia, provides annual housing price indices for both owned and rental housing nationally and for 30 cities back to 1890. These series are based on extensive work using archival newspaper real estate sections. In particular, over 2.7 million listings were digitised, with information on listed price, size, type, and location within each city. We use modern methods of index construction on these city-level datasets to generate like-for-like comparisons over time and then aggregate up to provide new national indices. These series are unique because they are constructed consistently across the period and reflect market conditions in each city at an annual frequency using publicly available sources.
In a new paper (Lyons et al. 2024), we introduce and explore this new dataset. We start by comparing the new HHP series to existing information on US housing prices. In both the cases of owned and market rental housing, we find that prices increased by more than previously thought. This is summarised in Figure 1, which plots indices of sale and rental prices for the US that adjust for inflation and where 1914=1.
Figure 1 Inflation-adjusted indices of US housing prices, 1890-2006 (1914=1)
Source: Analysis of Historical Housing Prices database, combined with Shiller (2015), BLS data and Rees-Jacob (1961).
Compared to the previous benchmark, we find roughly twice as much growth after inflation in the value of housing over this period (191% growth vs 101% growth in the Shiller index), with prices nearly three times higher in 2006 than in 1914. The difference is driven by the period 1950-1980, where our series shows 52% growth after inflation as opposed to 10%. This finding is in line with the Greenlees (1982) critique of the measure of housing prices used for this period, namely conforming home loans insured by the Federal Housing Administration (see also Davis et al. 2007). Further, we find a new and substantial housing market boom and bust in the interwar years corresponding to the Roaring Twenties and the Great Depression that aligns with narrative histories of this period.
Similarly, the new rent series shows far greater gains over the period 1914-2006. Rather than halving in real terms as indicated by the RoPR series, we find that rents show a mild upward trend with a clear cyclical component: inflation-adjusted market rents were in 2006 about one quarter higher than in 1914. In the scale and timing of its divergence, the new national series is consistent with an established literature that has highlighted limitations of the RoPR measure, especially from the 1940s to the 1980s, given the methodology used (Gordon and van Goethem 2007). It also broadly aligns with revisions to that series proposed by Crone et al. (2010).
The HHP sale and rental series can be used to generate new estimates of the returns to housing over the long-run, covering both capital gains and rental yield, nationally and at the city level. In Figure 2, we show the overall return to housing, by year, distinguishing between rental yield and capital gains. There is a clear difference between the two components. Rental income is less volatile but also gradually declining over time. Capital gains on the other hand are significantly more volatile but were the dominant element of returns to housing from the late 1960s until the end of our sample.
Figure 2 Nominal return to housing, by year and component, 1890-2006
Source: Lyons et al (2024).
In our paper, we explore not only the returns to housing but also the link between housing and the business cycle and the degree to which patterns of price growth across cities are linked to restrictions, both geographic and policy-based, to housing supply. We also outline an implication of these new results on estimates of the cost of living (and thus standard of living) over time. Given the importance of shelter in the basket of services in the Consumer Price Index (CPI), where market rents increased by substantially more 1914-2006 in our measure (a factor of 25, rather than 11), this means that inflation may have been understated.
Specifically, the CPI excluding shelter increased by an average of 3.3% per year 1914-2006, but the existing RoPR index increased by only 2.6% per year. In the HHP rental index, nominal rents grew by 3.5% per year, with the higher growth driven by the years after both World Wars as well as the 1965–1985 period. A back-of-the-envelope CPI using the HHP series as the housing component would yield a noticeably different estimate of overall inflation during the 20th century. It suggests that overall prices grew by 3.3% per year on average, rather than 3.1% per year, 1914–2006, and thus that there has been less of an increase in living standards in the US since WWI than previously thought.
We believe that this new dataset is a helpful resource that can be used to answer important questions of relevance to policymakers, for example on bubbles in housing, on spatial patterns in housing price movements over the long run, on the factors affecting prices city by city, and on the affordability of housing.
Source : VOXeu
