Amazon HQ Split: Housing Impact Should Be Diluted

Aaron Terrazas
Updated

Updated Nov. 13, 2018: The fact that Amazon chose two of the country's priciest housing and labor markets suggests that the decision to look beyond its original Seattle home is less about low costs and more about room to grow. Both Washington, D.C., and New York have similar housing costs to Seattle, but have historically been quicker to build upward and outward to meet the housing needs of newcomers. While 25,000 new jobs is no joke for any region, they should be able to accommodate this growth. They also offer much more developed public transit infrastructure, deeper labor markets, and better air accessibility. Given their size and more responsive housing supply, locals should not necessarily expect the same surge in rents and home values that Seattle has witnessed alongside its own Amazon boom.

Editor’s note: This research initially was published on April 25, 2018.

  • If Amazon workers start arriving one year from now, Nashville could see the pace of rent appreciation more than double from what Zillow expects without Amazon HQ2.

  • Denver and Los Angeles would also see a meaningful bump to rent growth from the estimated 50,000 workers Amazon plans to hire.

  • Most markets analyzed are not expected to see as large a bump in rents attributable to Amazon as that experienced in Seattle.

Later this year, Amazon is expected to name the city it has selected for its second headquarters (“Amazon HQ2”) – a sweepstakes of sorts that brings the prize, opportunity, and potential pain, of more than 50,000 highly paid jobs to the winning region.

And with those new jobs will likely come new residents, which in turn will boost demand for housing and potentially push up housing costs, particularly rents.

Extrapolating from the experience of how local rents have responded to the arrival of workers in recent years across each of the 17 metro markets that Amazon is considering for its second headquarters,[1] we estimate that Nashville, Denver and Los Angeles could experience the largest boost to local rents on top of that already anticipated. Toronto, Chicago and Indianapolis could see the most modest extra bumps to annual rent appreciation.

A panel of more than 100 housing experts recently surveyed by Zillow ranked Atlanta and Northern Virginia as the two most likely cities to be chosen for Amazon HQ2. Both can expect a relatively modest boost to rents if selected – an additional 0.4 percentage point boost to rent in Atlanta, and a 0.6 percentage point boost in the Washington, D.C., area which includes Northern Virginia.

Over the past decade, Amazon's growth from a startup bookseller into one of the nation's leading retailers has transformed its original hometown of Seattle. The region has consistently ranked as one of the nation's hottest housing markets, though many forces beyond Amazon's growth have propelled the Seattle market in recent years.

All but two of the HQ2 finalist markets should expect a smaller impact on rents than Seattle has experienced.[2]

In many respects, the results are intuitive. Smaller markets and/or markets with a record of less responsive (or "elastic," in economist-speak) housing supply should expect a larger increase in rent growth. Bigger markets and/or those with a history of rapidly expanding the local housing supply should expect a smaller impact. And in larger markets with more diversified labor forces, fewer workers would likely have to relocate to the area to fill the new jobs, which is also likely to blunt rent growth.

But even in areas where we expect only small bumps in rent appreciation, it does not necessarily follow that there will be no impact on the housing market overall. In Indianapolis, for example, there is likely to be essentially no boost to rent appreciation in part because owner-occupied homes are so readily available, affordable and historically responsive to new demand. So while local rent appreciation might remain muted, Indy-area home sellers and builders should reasonably expect to be fairly busy if their city is chosen to be home to Amazon HQ2.

Amazon HQ2: Past is Prologue

Underlying these estimates is the assumption that the newcomers ("worker inflow") associated with Amazon HQ2 will be similar to newcomers to the metro over the past half-decade in their decision to rent or buy. This analysis also ignores second-order effects, such as worker inflows associated with business clusters that may pop up around Amazon's campus (as they have in Seattle).

We also assume that each community's ability to add new units in response to changes in demand will be similar to its recent past. In markets where newly arrived workers have skewed toward highly-educated, higher-wage sectors in recent years (such as in Boston or Washington, D.C.), these are probably reasonable assumptions. But in markets where there may be more of a skill or wage gap between recent arrivals and the still-to-come workers moving in because of Amazon, these assumptions may be more of a stretch.

Put another way: The engineers arriving in Boston for a potential Amazon job may not be too different in terms of wage expectations and education from the engineers who moved there recently for a job at General Electric's Boston headquarters. So we can more reasonably assume the future housing market in Boston will respond in a similar way to Amazon's presence as it has to other recent instances of worker inflow. But assembly line workers that arrived to Indianapolis several years ago to work a manufacturing job at Carrier Corp. may be quite different from a future Amazon engineer, so it's more difficult to model the impact these newcomers may have on the local housing market.

Methodology

To estimate how rents might respond to worker inflows associated with Amazon's second headquarters, we compiled:

We then estimated six different models of how rents have responded to inflows of workers from outside the metro, scaled by total metro employment in the previous quarter. We scale worker inflows since the arrival of a given number of workers will represent a larger shock to relative demand in a smaller metro than in a larger metro. The six models include:

  • The annual percent change in rents as a function of worker inflows, scaled by total employment

  • The quarterly percent change in rents as a function of worker inflows, scaled by total employment

  • The annual percent change in rents as a function of the four-quarter change in worker inflows, scaled by total employment

  • The quarterly percent change in rents as a function of the one-quarter change in worker inflows, scaled by total employment

  • The change in the annual percent change in rents as a function of the four-quarter change in worker inflows, scaled by total employment

  • The change in the quarterly percent change in rents as a function of the one-quarter change in worker inflows, scaled by total employment.

The models are estimated over the period for which data are available for all three data series (Q4 2010 through Q3 2016).

To estimate quarterly inflows associated with Amazon's HQ2, we divide the 50,000 total workers Amazon has said it plans to ultimately hire by a "local hire factor." Presumably, in larger labor markets, a smaller share of workers will have to move to the area – instead, Amazon can hire from the existing labor pool.

  • For metros with a current employment base of more than 3 million, we assume that 60 percent of total hires will be local hires.

  • For metros with total employment between 2 million and 3 million, we assume that 50 percent of total hires will be local hires.

  • For metros with an employment base between 1 million and 2 million, we assume that 40 percent will be local hires.

  • For metros with an employment base under 1 million, we assume 30 percent will be local hires.

While we have no empirical basis for these factors, informal conversations with tech industry recruiters suggest they are reasonable.

We also assume it will take Amazon approximately 7.5 years to hire all 50,000 promised workers – roughly the length of time it took for Amazon to hire up at its Seattle headquarters – and that hiring will occur at a constant pace over that period. If staffing initially ramps up at a faster pace, the boost to rent appreciation could be hire during the early years and then taper as the pace of hiring slows.

We have a series of five criteria among the six models, starting with the most rigorous set of selection criteria to identify the strongest model predictions, then gradually relaxing the selection criteria until we have an estimate for all 16 U.S. metro markets. As a result, we believe that some of the predictions are stronger than others.

  • Five of the metro predictions passed the most rigorous selection criteria: a limited subset of models, a p-value below 0.1, and a maximized adjusted R-squared.

  • Six of the metro predictions were selected with slightly less rigorous selection criteria: a limited subset of models, parameter sign restrictions, a p-value below 0.25, and a maximized adjusted R-squared.

  • One metro was selected with a less rigorous selection criterion: an expanded set of models, parameter sign restrictions, a p-value below 0.1, and a maximized adjusted R-squared.

  • One metro was selected with an even less rigorous selection criterion: an expanded set of models, parameter sign restrictions, an adjusted R-squared greater than zero, and a minimized p-value.

  • The remaining four metros were selected with the least rigorous selection criteria: an expanded set of models, parameter sign restrictions, and an averaging of estimates over all models.

While the results are broadly intuitive, we consider the predictions selected using the first three selection criteria to be the most credible.

Due to data limitations, the estimate is slightly different for the Toronto market. We rely on annual rent data from the Canada Mortgage and Housing Corporation's (CMHC) Housing Observer, which includes both proprietary CMHC data and data compiled from Statistics Canada. We estimate a model of annual percent changes in Greater Toronto rents as a function of Greater Toronto population growth over the period from 2003 to 2016. Using similar assumptions about the magnitude and pace of worker inflows associated with Amazon's HQ2 as we applied to the U.S. markets, we then estimated how Toronto rents might respond.

 

[1] The 20 finalist cities cover 17 metro areas since three finalist cities are in the Washington, DC metro area and two are in the New York metro market.

[2] To validate these results we also computed the hypothetical effect of similar worker inflows on Seattle metro rents. The resulting estimate of a 2.1 percentage point boost to annual rent appreciation is slightly higher than what we have previously estimated using a different methodology. Some of the difference might be explained by timing: Our earlier analysis estimated how Seattle rents responded to the jobs boom in and around Amazon's headquarters between 2011 and 2015 while the estimates of the models applied here show how Seattle rents might be expected to respond to an additional 50,000 new workers starting a year from now.

The post Amazon HQ Split: Housing Impact Should Be Diluted appeared first on Zillow Research.