We use cookies on our website to give you the best online experience. Please know that if you continue to browse on our site, you agree to this use. You can always block or disable cookies using your browser settings. To find out more, please review our privacy policy.

About


Real Estate Research provided analysis of topical research and current issues in the fields of housing and real estate economics. Authors for the blog included the Atlanta Fed's Jessica Dill, Kristopher Gerardi, Carl Hudson, and analysts, as well as the Boston Fed's Christopher Foote and Paul Willen.

In December 2020, content from Real Estate Research became part of Policy Hub. Future articles will be released in Policy Hub: Macroblog.

Comment Standards:
Comments are moderated and will not appear until the moderator has approved them.

Please submit appropriate comments. Inappropriate comments include content that is abusive, harassing, or threatening; obscene, vulgar, or profane; an attack of a personal nature; or overtly political.

In addition, no off-topic remarks or spam is permitted.

November 18, 2014

Can the Atlanta Fed Construction and Real Estate Survey Predict Home Sales?

The slow recovery in housing remains an item of note in statements from the Financial Open Market Committee. That it's still something of a concern means that many people pay attention to housing-related data releases, several of which are due out this week, because they can shed some light on the direction of housing and the economy. The builder confidence index, released today, got things off to a good start by showing a four-point rise, from 54 to 58 (values greater than 50 mean that more builders view conditions as good rather than as poor). House starts and existing sales are due Wednesday and Thursday, respectively.

At the Atlanta Fed, we conduct a monthly survey of regional builders and real estate brokers to get their perspectives of the market. In August, we began to look at the results a little differently to see if they could tell us anything about subsequent housing-market data releases. In that exercise, we investigated the correlation between the expectations of our homebuilder contacts for construction activity and subsequent housing starts. We found that our builders are on point, more or less, and we reported on that discovery in an August post. We recently repeated the exercise, this time to explore the predictive power of the outlook for home sales of our homebuilders and residential brokers for subsequent new and existing home sales data releases. We report on our findings in this post.

Brokers and builders expect new home sales to rise
The September home sales data showed us that existing single-family home sales increased by 1.9 percent from the year-earlier level and new home sales increased by 22.6 percent. This news is fairly consistent with the reports we received from our real estate business contacts about September sales activity; more brokers and builders noted an increase than noted a decrease in home sales activity from the year-ago level.

But what exactly did our survey respondents tell us about their outlook for home sales? Diving deeper into the data, we find that brokers' and builders' outlooks remain mildly positive and that the two groups have tracked each other fairly closely in recent years. (In the pre-2011 period, brokers and builders diverged more sharply.) Specifically:

  • Of builder respondents, 40.0 percent indicated that they expect new home sales to increase over the next three months, 32.0 percent expect activity to decline, and 28.0 percent expect home sales activity to remain about the same. The home sales outlook diffusion index value for builders was 0.08.
  • Of broker respondents, 22.5 percent indicated that they expect new home sales to increase over the next three months, 27.5 percent expect activity to decline, and 50.0 percent expect home sales activity to remain about the same. The home sales outlook diffusion index value for brokers was -0.05.

SE-Home-Sales

The chart below features two scatter plots of the diffusion index value for the broker and builder home sales outlook on the horizontal axis and the year-over-year change in the three-month moving average of single-family home sales (for Alabama, Florida, Georgia, Louisiana, Mississippi, and Tennessee) on the vertical axis. Given that we are asking contacts to be forward-looking, we lag the contact responses.

Broker-Home-Sales

Do home sales expectations correlate with subsequent sales data?

Three things stand out on this chart. First, if builders and especially brokers (who tend to be an optimistic lot) predict a decline, the subsequent home sales data release will probably be poor. Only a modest bit of net optimism is of little comfort—some of the worst declines occurred in years with net positive (albeit small net positive) outlooks.

Second, for builders, if the index is greater than 0.3, we find that sales generally grow—except for between August 2012 and April 2013, when sales did not match builders' optimism. When the broker index is above 0.3, sales either grow or decline by a smaller amount than when the index is negative. Like the builders, the broker panel missed the sales declines from August 2012 and April 2013. The brokers also missed the declining real estate market in 2006 to early 2007 (see the green triangles in the chart above)—despite a declining market, the broker index remained lofty until May 2007.

Third, the official statistics on housing sales could go either way when index values are between ‑0.1 and 0.3. This shouldn't come as a complete surprise, particularly because a diffusion index value near zero (regardless of whether that value is positive or negative) indicates that responses from contacts were mixed. And as we can see in the scatter plot above, large declines were much more likely given the time period covered.

A simple regression indicates that the outlook could explain just under 50 percent of the variation in sales measure, which indicates that our poll does a decent job of predicting subsequent sales. Given this finding, what do we now expect home sales to look like? The most recent downward trend in respondents' outlook puts the diffusion index in the center, suggesting that declines in seasonally adjusted sales over the next several months are just as likely as increases in sales.

The poll was conducted October 6–15, 2014. Sixty-five business contacts across the Southeast participated (40 residential brokers and 25 builders). To explore the latest poll results in more detail, please visit our Construction and Real Estate Survey page.

Photo of Carl Hudson By Carl Hudson, director for the Center for Real Estate Analytics in the Atlanta Fed´s research department, and

 

Photo of Jessica DillJessica Dill, senior economic research analyst in the Atlanta Fed's research department

September 18, 2014

The Economic Effects of Urban Renewal

Editor's note: An earlier version of this post inadvertently included a paragraph from last week's post. The corrected post is below, and we apologize for the oversight.

This year, the 50th anniversary of the "War on Poverty," has seen an effort in the news media and among policy commentators to review the success and failure of past efforts to address poverty (see, for example, this, this, this, and this). Some of these efforts have included place-based policies such as the Model Cities program, which attempted to improve housing stock and reduce urban blight at the neighborhood level. In part, this renewed interest is policy-relevant: many cities are struggling with blight in the wake of the foreclosure crisis, and place-based policy has returned to popularity. For these reasons and more, I was quite interested to read a recent article in the American Economic Journal: Applied Economics. "Slum Clearance and Urban Renewal in the United States" by William J. Collins and Katherine L. Shester revisits the topic of urban renewal programs in the latter part of the last century.

The set of policies loosely referred to as "urban renewal" has been controversial since implementation. In fact, the programs changed a lot from 1950 to 1974, largely in reaction to the outraged response and perceived failures of early efforts. Title I of the 1949 Housing Act, which focused on "slum clearance," was a precursor to the 1954 Housing Act, which shifted the emphasis away from demolition and towards rehabilitation and preservation. Later legislation added programs to smooth the relocation process for those who were displaced by Title I programs and to direct resources towards the elderly poor. Throughout the 1960s, policy shifted away from changing the quality of housing stock towards creating a suite of policies focused on healthy communities. In 1965, as a result of a major reorganization, the Housing and Home Finance Agency, which had administered Title I, became the Department of Housing and Urban Development, commonly known as HUD. Finally, in 1968, the Fair Housing Act passed, further affecting the dispersal of funds.

In the early sixties, Jane Jacobs was one of the more famous critics of the destruction of historic neighborhoods and reconstruction along rationalist, modernist lines. In her 1961 classic, The Death and Life of Great American Cities, she argued that cities embodied organized complexity and that so-called "disorderly" slums were better than the rationally planned spaces that displaced them, both economically and socially. Other research on urban renewal has focused on political, social, and legal implications. This line of inquiry has focused on the impact of eminent domain on property rights, aesthetic concerns about how to incorporate historic preservation into revitalization, and concerns of justice and equity, primarily the issue that urban renewal placed the burden of displacement and disruption onto poor and minority residents without due consultation or compensation (see Gans 1962, Gotham 2001, Jacobs 1961).

The 2013 Collins and Shester paper cites this literature, but is distinct from it in its quantitative, nationwide study of economic impacts. It evaluates the effect of a series of programs over a 30-year period across 458 cities, and calculates that effect on broad economic outcomes. The authors measure urban renewal by combining the dollars allocated under the various programs implemented between 1950 and 1974. They evaluate the combined effect of these programs using a regression model. This model estimates the impact of federal dollars spent on the change in economic health of each city between 1950 and 1980. Using census-region fixed effects, the authors evaluate the impact of expenditures on median income, median property values, the employment rate, and the percentage of people living in poverty.

The authors' first-stage findings show that federal dollars spent on urban renewal projects between 1950 and 1974 had a negative effect on various economic outcomes. However, Collins and Shester suspect there is endogeneity in the relationship they are trying to uncover. That is, they say we cannot be sure what causes what: did urban renewal cause economic growth or decline, or did blighted cities pursue more urban renewal? In the latter case, even if the program improved the economy, these cities might still be doing more poorly than cities that had no blight to begin with.

The authors deal with endogeneity using an instrumental variable approach. That is, they seek to use exogenous variation in the allocation of federal funds. The variable they use is the year in which a state passed enabling legislation that made these sorts of projects legal. At first glance, this isn't a great instrument. Instrumental variables have to meet what's called the "exclusion restriction" to be credible. That restriction is untestable; you have to evaluate this claim on its merits. So, for us to believe this instrument delivers credible result, we have to be convinced that a state's decision to pass enabling legislation affects economic outcomes only by the way it influences urban renewal expenditures. There can't be any other chain of effects of related issues that connect those two events—the instrument and the outcome.

Collins and Shester perform several tests to justify their instrument. First, they look just at the effect of the instrument in places where court cases affected the timing of the laws passing. Then they perform a test of known effects to see whether their model predicts the economic growth in rural areas where urban renewal was not pursued. Finally, they use an alternate specification of the instrument. The instrument holds up under these examinations.

The authors then use their metric to predict the urban renewal funds distributed, and then use that predicted value in the original model. In this specification, urban renewal dollars have a strong positive effect on income and property values. These findings are consistent across several specifications and robustness checks. Furthermore, they find no effect on employment or poverty rates, leading them to posit that the positive effects they observe were not generated by displacement of poorer residents from inner cities. As a whole, these results suggest that overall, urban renewal programs created positive growth in average wages and property values.

A concern is that these conclusions rest on the credibility of the instrumental variable, and I'm not sure that the instrumental variable meets the exclusion restriction. I also wonder whether the average effects might reflect underlying variation in the effect of individual programs in urban renewal as well as different contexts where the program was applied. A map of the instrument (below) shows a strong spatial component to the instrument. Of the 458 cities that the authors measured in 1950–80, 68 percent of the cities, or 311, were in states that passed enabling legislation immediately. Regions in the Northeast, Midwest, and West pursued urban renewal programs immediately. These states were the most industrialized parts of the country; they experienced sectoral change and decline of their manufacturing center. The more agricultural, conservative areas of the country pursued funds relatively later, and received funds under later programs.

Instrumental-variable-map
Source: Collins and Shester 2014, author's calculations

This makes me wonder if there isn't sufficient variation in the manufacturing states, and that the instrumental variable instead down weights these cases, providing in essence a regional estimate. Looking at the first stage results within each census region, we find that the results vary by region. For heavily industrial regions—the Mid-Atlantic, East North Central, and East South Central—urban renewal funding had a negative on growth. The other regions show a positive relationship between urban renewal and growth and economic growth.

There is also inconsistency in the second-stage, or instrumented, results within each region. The two regions in the Midwest, stretching from Wisconsin to New York, drop out as there is no variation. The regions on the eastern half of the nation show a positive effect, while those in the West show a negative effect.

Collins and Shester want to evaluate the treatment effect of urban renewal dollars by creating as-if-random variation in the administration of urban renewal funds. But if we aren't convinced that the instrument meets the exclusion restriction, or that the policy is having a constant effect, then what can we make of the results generated by this instrumental variable? We might surmise that the instrument is telling us something about the impact of the program in the subset of cities where the instrumental variable generates variation. If we believe that the study design can actually capture the effects of urban renewal, we might think of these new estimates as telling us the average effect of later urban renewal projects in 158 cities in the South and rural West, and not so much the effect of the program in the 311 cities where urban renewal was most intensively pursued.

Photo of Elora RaymondBy Elora Raymond, graduate research assistant, Center for Real Estate Analytics in the Atlanta Fed's research department, and doctoral student, School of City and Regional Planning at Georgia Institute of Technology

August 29, 2014

Real Estate Business Contacts on Target

After several months of year-over-year declines in housing starts and increasing concern that momentum in the housing market was slowing, we received some positive news this past week. The U.S. Census Bureau and Department of Housing and Urban Development, or HUD, jointly released the July 2014 construction statistics: total housing starts were up 15.7 percent from the upwardly revised June estimate and were 10.7 percent higher than the year-earlier level.

This news is fairly consistent with the reports we received from our real estate business contacts. Each month since 2006, the Atlanta Fed has conducted a poll of brokers and builders from six states in the Southeast in an effort to gather anecdotal intelligence. (These states are Alabama, Florida, Georgia, Louisiana, Mississippi, and Tennessee.) This intelligence not only helps form the basis for the Atlanta Fed’s construction and real estate submission to the Beige Book, but it is also considered a helpful tool for collecting insight on current market conditions before the release of the official housing statistics.

Our August poll results came in last week. Altogether, 60 business contacts participated—44 of them were residential brokers and 16, builders. The results reflect activity in July 2014. What did we find? As it relates to housing starts:

  • Sixty-three percent of respondents indicated that current construction activity was ahead of the year-earlier level, 19 percent said construction activity was down, and the rest indicated that activity was about the same. In August 2014, the current construction activity diffusion index value was 0.44.
  • Fifty-seven percent of respondents indicated that they expect construction activity to increase going forward (specifically, over the next three months), 13 percent said they expect activity to decline, and 31 percent said they expect the level of construction activity to remain about the same. The construction activity outlook diffusion index value also happened to be 0.44.

Southeast-construction-activity

We compute a diffusion index of responses to help us more easily track the trend over time. Basically, we subtract the share of respondents reporting decreasing activity from the share of respondents reporting increasing activity to arrive at a diffusion index value. (We do not factor those reporting no change in activity into the diffusion index equation.)

We tend to think of any diffusion index value higher than zero as an indication that construction activity is increasing, so we’d like to think that our signal from real estate business contacts was somewhat accurate. Using this same rule of thumb, though, it is hard not to notice that our construction activity diffusion indexes have been tracking above zero since 2011, and this does not necessarily jibe with recent data releases.

So how useful is the Atlanta Fed’s Southeast housing market poll? Well, we’ve spent some time evaluating survey data against the actual outcomes to better understand what we can reasonably take away from the results. We ask contacts what their expectation is for construction activity next three months versus the year-ago period. The chart below features a scatter plot of the diffusion index on the horizontal axis and the year-over-year change in the three-month moving average of single-family housing starts on the vertical axis. (Given that we are asking them to be forward-looking, we lag the contact responses.)

Builder-contact-poll

We found that when the diffusion index value is greater than 0.3, starts subsequently grew. When the diffusion index values falls below -0.3, starts tended to fall. But we found that between -0.3 and 0.3, the official statistics on housing starts could go either way. This shouldn’t come as a complete surprise, particularly because a diffusion index value near zero (regardless of whether it is positive or negative) indicates that responses from contacts were mixed. Real estate markets are in fact local, so it doesn’t mean that our contacts were wrong. It simply means that their responses are less likely to match the larger picture when aggregated. And as we can see in the scatter plot above, large declines were much more likely given the time period covered.

Using a simple regression model, we would have expected June 2014 starts in the six states we cover to be around 114,000 rather than the roughly 103,000 that were reported. So while the poll results may not serve as a perfect early warning system, they do correlate with subsequent starts in the states we cover. To explore the poll results in more detail, please visit the Atlanta Fed's Construction and Real Estate Survey page.

Photo of Carl Hudson By Carl Hudson, director for the Center for Real Estate Analytics in the Atlanta Fed´s research department, and


Photo of Jessica DillJessica Dill, senior economic research analyst in the Atlanta Fed's research department

March 27, 2014

Limiting Property Tax Assessments to Slow Gentrification

A recent New York Times article on gentrification discussed a number of cities—including Boston, Philadelphia, and Washington, D.C.—that are planning to freeze property tax assessments for long-time homeowners in gentrifying neighborhoods. The concern is that rising house prices will also raise property assessments, forcing low-income residents to move to escape the greater tax burden and thereby accelerating the pace of gentrification. Although the desire to protect existing residents from gentrification appears to be new, laws capping assessment growth for all property or all primary homes ("homesteads") have been around since Californians passed Proposition 13 in 1978. After California, a number of additional states passed laws limiting how quickly an individual property's assessed value could increase. The bulk of these laws passed in the early eighties to the mid-nineties, and advocates for the law were concerned, at least in part, with limiting the size of local government. If this tax backlash of the previous decades is uncorrelated with more recent gentrification pressures, this may be a good test of statewide assessments caps.

Using a data set of low-income central-city neighborhoods that Dan Hartley of the Cleveland Fed assembled from the 2000 census and the 2007 American Community Survey, we can look at the share of neighborhoods that gentrified in capped and uncapped states. Hartley shows that a central city moving from below-median-MSA house price to above-median house price is a good indicator of gentrification. Relying on the table of statewide assessment caps that Haveman and Sexton compiled, we identify 10 states and the District of Columbia (plus the city of New York) with the strictest limits. In these states, assessed value can increase only at the rate of inflation or by a fixed percentage ranging from 2 percent (California) to 10 percent (Texas). Table 1 presents the share of neighborhoods that gentrified in capped and uncapped states.

Table 1: Share of Neighborhoods that Gentrified between 2000 and 2007

Note that neighborhoods protected by assessment caps actually gentrified faster than those in states without them.

However, we might worry that the decision to impose statewide assessment caps was not random. In the case of Prop 13, rising home prices was certainly a factor in rising property taxes. It is possible that some underlying factor may drive statewide price up but also cause poor inner-city neighborhoods to appreciate faster than other homes in the metro area. One candidate is restrictive zoning laws that limit densification of already desirable neighborhoods. Such laws could both drive up aggregate house prices and push homebuyers into more marginal neighborhoods, causing them to appreciate relatively faster. However, assessment caps are only one possible response to rising property taxes. If voters wish to limit the growth in property taxes, they don't need capped assessments—they can restrict the growth in property tax revenues directly. At the same time, assessment caps that don't also cap the property tax rate don't actually constrain property taxes, but instead shift the tax burden from longtime owners to new buyers. In Table 2, we limit the sample to states that have a binding revenue growth cap or that jointly cap assessments and municipal tax rates. In this case, we assume that, conditional on imposing a tax expenditure limit, the decision to cap assessments rather than property tax revenue is random. We rely on the work by Hoyt, Coomes, and Biehl (2011) to identify various statewide tax expenditure limits.

Table 2: Share of Neighborhoods that Gentrified between 2000 and 2007 with some form of binding property tax limit

Limiting the sample to states that have chosen to constrain the property tax in some way, we still observe assessment caps seeming to accelerate gentrification rather than slow it. How can that be? One possibility is that because these are state-wide limits, the caps have reduced the turnover in more desirable neighborhoods, driving new homebuyers to marginal central-city neighborhoods. In that case, targeted assessment caps that apply only to currently low-priced neighborhoods could still be efficacious. On the other hand, the existence of an assessment cap may increase the long-run return from "pioneering" in a low-priced neighborhood.

So far, we have been using change in relative house prices as our definition of gentrification. However, advocates for assessment caps are plainly concerned about the ability of homeowners to stay in their home in the face of rising home values. While the in-migration of higher-income residents and house prices are highly correlated, we do not observe the duration of time that existing residents remain in their home. Unfortunately, there are few individual-level data sets with sufficiently granular geography to allow such an analysis. As an alternative, we can look at the change in median income of residents. This value is available at the census-tract level in the 2000 census and the 2007 American Community Survey. Table 3 presents change in median income for all census tracts and for gentrifying tracts with and without assessment caps. While median incomes rose in gentrifying neighborhoods (even as they declined nationally), they rose faster in tracts subject to an assessment cap. However, this difference is not statistically different from zero (p value 0.303).

Change in real income for gentrifying neighborhoods with and without assessment caps

Finally, assessment caps do nothing for renters, who may be impacted much more immediately by rising neighborhood quality than homeowners. It is possible that assessment caps could still allow a small share of long-time owners to stay, and the observed effects are just dominated by the movement of renters. If we had access to administrative data with finer geographic identifiers, we could look at whether neighborhoods that gentrified with assessment caps now exhibit more income or racial heterogeneity than neighborhoods without. However, looking only at aggregate data, property taxes do not appear to be a primary driver of neighborhood change, and concerns about gentrification do not appear to warrant interfering with the assessment process.

Photo of Chris CunninghamChris Cunningham, research economist and assistant policy adviser at the Federal Reserve Bank of Atlanta