We've heard a fair amount of discussion lately about the decline in the first-time home buyer share of market, both from our business contacts in the homebuilding and real estate industries and from various media outlets. One example is the July 22 Wall Street Journal (WSJ) article that stated:
First-time home buyers, long a key underpinning of the housing market, are increasingly getting left behind in the real-estate recovery. Such buyers, typically couples in their late 20s or early 30s, have accounted for about 30% of home sales over the past year. They represented 40% of sales, on average, over the past 30 years, and accounted for more than 50% in 2009, when recession-era tax credits fueled the first-time market, according to data from the National Association of Realtors. The depressed level of first-time buyers could prove to be a drag on the housing rebound and the broader economic recovery over the longer haul.
Other media examples include a second WSJ article that ran on August 27, a Bloomberg Businessweek article that ran on August 7, and a USA Today article that ran on June 29. All articles cite data publicly available from the National Association of Realtors® (NAR).
When we took a closer look at these articles, we found that they actually intermingle two different sources of data, both from the NAR. The "30%" share of first-time home buyers over the past year comes from the Realtors Confidence Index, a monthly survey of approximately 3,500 Realtors that runs from October 2008 through July 2013. The "40%" average comes from a second NAR source, the Profile of Home Buyers and Sellers, an annual survey of approximately 8,500 households that runs from 2001 through 2012. We’ve graphed both time series in chart 1 below.
Not surprisingly, the two series do not completely align because each series was constructed with a different methodology. Given the different methodologies, it is not appropriate to use data from the annual NAR survey as a reference point for data from the monthly NAR survey.
With that said, the central question still lingers: has there in fact been a decline in the share of first-time buyers? Instead of comparing levels across data series, we would suggest a comparison of trends. To provide a broader perspective on the trend in the first-time buyer share, we introduce two additional monthly measures: the Census Bureau's American Housing Survey Public Use Microdata and the Campbell/Inside Mortgage Finance HousingPulse Tracking Survey.
It's important for us to note the methodology of each data series before jumping into a comparison of the trends. As we point out in the table, while the data are all survey-based, they have varying frequencies, coverage periods, and types of respondents.
Source | Series Name | Frequency | Coverage | Respondents | ||||
Census Bureau | American Housing Survey–Public Use Microdata | Monthly (See the methodological endnote) | Oct 1983–Sept 2011 | Survey of 60,000 housing units reduced to owner-occupied households that were recent movers | ||||
Campbell/Inside Mortgage Finance | HousingPulse Tracking Survey | Monthly | July 2009–July 2013 | Survey of (+/-) 2,000 real estate agents | ||||
National Association of Realtors | Realtors Confidence Index | Monthly | Oct 2008–July 2013 | Survey of (+/-) 3,500 Realtors | ||||
National Association of Realtors | Profile of Home Buyers and Sellers | Annual | 2001–2012 | Survey of (+/-) 8,500 households |
As chart 2 shows, due to the differences in methodology, none of the series align perfectly. However, each series tells a similar story.
To get a better sense of the trend, we plotted each series individually (not shown). First, we examined the monthly American Housing Survey time series that we constructed (see the methodological endnote). The long-term linear trend line from October 1983 through September 2011 was slightly upward-sloping. Many have argued, though, that the first-time homebuyer tax credit program pulled demand forward and that the tax credit period (July 2008–September 2010) distorts the overall long-term trend. Indeed, when we exclude this time period, we find that the slope becomes slightly downward-sloping. We observe a similar trend when we fit a trend line to the National Association of Realtors' Profile of Home Buyers and Sellers time series. From 2001 through 2012, the trend is slightly upward-sloping when we include the tax credit period and slightly downwardly-sloping when we exclude the tax credit period.
When we take a closer look at the series with shorter time horizons, we generally find trends similar to those of the series with longer time horizons. The Campbell/Inside Mortgage Finance series clearly has a downward-sloping trend when we include the tax credit period and only a slightly downward-sloping trend when we exclude it. When the National Association of Realtors' Realtors Confidence Index is fitted with a trend line, the trend is clearly downward-sloping regardless of whether we include or exclude the tax credit period. We do think that it is important to point out that once we remove the tax credit period from these shorter term series, we're left with slightly less than three years’ worth of data. Given the shorter length of these series, we feel they are more likely to reflect shorter-term fluctuations than any shift in the longer-term trend.
With that said, it was striking to us how close to zero the slopes were of the slightly downward-sloping trend lines that exclude the tax credit period. When we calculate the trend of first-time buyer participation using an OLS regression and incorporate the tax credit period as a 0,1 dummy variable (which had a value of 1 if the tax credit was in effect, and 0 otherwise), we find that the slightly downward-sloping trends are not statistically different from zero after accounting for the effect of the tax credit.
In other words, we agree that the tax credit period distorts the trend, we think it is best to exclude it when interpreting the trend in first-time buyer share, and we interpret the trend in the first-time buyer share as flat across each data series when the tax credit period is excluded.
So to wrap up, we agree with the WSJ's statement that first-time buyers are a key underpinning of the housing market. However, we do not share the concern about weakness in housing demand going forward because we are not convinced that the data indicates a material decline in first-time buyer participation. Claims of a decline in first-time buyer participation that appear to be based on a comparison of data across different surveys should be treated with caution. There are several sources of data available for tracking the first-time buyer share of market. In comparing the trends of each series separately, we don’t find there to be much in the way of a material decline in the share of first-time home buyers over the time periods and data series we examined.
By Jessica Dill, senior economic research analyst, and
Ellyn Terry, an economic policy specialist, both in the Atlanta Fed's research department
(Return to table) | Methodological endnote: In order to get a longer-term view of the data, we constructed a monthly time series of first-time homebuyer share using the American Housing Survey (AHS) back through 1983. Our first-time buyer data point is derived from several questions in the AHS. First, we determine the current status of the person occupying the unit (PERSON module—Chart A). We then drill down further and look at the recent mover module to flag all respondents that moved residences since the prior survey year (RMOV module—Chart A).
Of those who indicated that they currently own their residence and indicated that they recently moved, we created flags for their status in their previous residence. The portion of respondents that indicated that they rented their previous residence or that they lived rent-free at their previous residence were combined to create our estimate of the first-time buyer share. We took this exercise one step further by utilizing the move date to construct a monthly time series of first-time buyer share dating back to 1983 from the biennial AHS dataset. While we did not weight the data, we did compare our numbers to yearly tables on the AHS website containing weighted data by survey year (as opposed to move date) and found our numbers to be comparable.