Tag Archives: Census Bureau

Census Bureau Remodeling Data Revisions Out of Sync with Other Market Indicators

Census Bureau Remodeling Data Revisions Out of Sync with Other Market Indicators

 by Abbe Will
Research Analyst
Since 2007, the Joint Center for Housing Studies has produced a quarterly Leading Indicator of Remodeling Activity (LIRA), which makes use of several economic indicators that historically have had strong correlations and leads over remodeling spending to anticipate near-term changes in the market. The Joint Center relies on the homeowner improvement expenditure data reported in the U.S. Census Bureau’s monthly Construction Spending Put in Place series (C-30) to estimate and benchmark its LIRA.
On July 1, the Census Bureau released its regularly scheduled annual revisions to the C-30, which adjusted the monthly improvement spending estimates back two calendar years to January 2011. As seen in Figure 1, these revisions increased estimates of home improvement spending by 6% for 2011 (from $114 billion to $121 billion) and decreased spending over 10% for 2012 (from $125 billion to $112 billion).


Source: US Census Bureau, Value of Private Construction Spending Put in Place (C-30).

Typically, these annual revisions are minimal and, in the past, changes were always in the same direction as the original estimates, often revising the whole series downward somewhat (Figure 2). This time, not only was the magnitude of the revisions significantly larger than in recent years, but the direction of the revisions was extremely divergent from what could have been expected based on previous annual revisions.


Source: US Census Bureau, Value of Private Construction Spending Put in Place (C-30).

Certainly, the extent of these revisions by itself calls for a thorough analysis and understanding of the reasons behind such dramatic changes, but the fact that the adjustments are at odds with other key industry data is even more worrisome. Figure 3 compares the annual rates of change in data series that historically correlate highly with home improvement spending (and are used as main inputs in the LIRA) with the pre- and post-revision C-30 data. As seen in the figure, key industry indicators including retail sales at building material and supply stores, remodeling contractor sentiment (RMI), pending home sales (PHSI) and single family housing starts all trended closely with the pre-revised C-30 estimates of homeowner improvement spending since January 2011.

Sources: US Census Bureau, Value of Private Construction Spending Put in Place (C-30), Monthly Retail Trade Report and New Privately Owned Housing Units Started; National Association of Home Builders Remodeling Market Index (RMI); and National Association of Realtors© Pending Home Sales Index (PHSI).
At this time, there is no obvious explanation for why the revisions to the C-30 improvements data were so extreme this year. As part of the Joint Center’s investigation of this issue, we will be in contact with the federal agencies involved in collecting the survey data and developing these estimates to assess whether changes in survey methodology or weighting procedures, for example, might explain these large shifts. As the Joint Center reviews the underlying source data for home improvement spending and the procedures that generate these estimates, we have decided to forgo publication of the LIRA this quarter. The next LIRA is scheduled to be released on October 17, 2013.


Leave a comment

Filed under JCHS, LIRA

JCHS: A Word of Caution about Census Bureau Projections

A Word of Caution about Census Bureau Projections

 by George Masnick
The Census Bureau recently released its high and low immigration series population projections going out to 2060, complementing its middle series released in December 2012.  Among the talking points in the press release announcing the projections was the assertion that the high immigration series “projects that the U.S. resident population will become majority-minority by 2041, two years earlier than the December (middle series) projection of 2043.”The point is well taken that the growth of the minority population depends upon future levels of immigration, and higher immigration means an earlier date at which the country becomes majority-minority. But between now and the 2040s there is a great deal more uncertainty about immigration trends than is captured in the Census Bureau’s latest assumptions, including uncertainty about how attitudes and norms will affect who is counted as minority.  In addition, all three of the Census Bureau’s population projections use the same assumption about projected fertility levels of each race/Hispanic origin group.  Not only are fertility trends difficult to predict over many decades, but different immigration levels will surely affect fertility rates.

The new middle series immigration assumptions trend from 725,000 per year in 2012 to 1.2 million in 2050. Low assumptions trend from 702,000 to 808,000, and high assumptions from 747,000 to 1.6 million annually.  All three immigration assumptions are well below those of the previous Census Bureau population projections released in 2008 and 2009, with the new high immigration series even projecting fewer immigrants in 2050 than the previous low immigration assumption (Figure 1). Such a wide range of uncertainty about future levels of immigration should make one wary about the reliability of projections that reach almost 50 years into the future.

Source: U.S. Census Bureau 2009 National Population Projections

However, the projected level of immigration is not the only area of uncertainty that will affect the projected white-minority tipping point.  Predicting a specific year when the population actually becomes majority-minority sometime three decades ahead will also require assumptions about how Americans in the future will identify themselves in terms of race and ancestry.

Perhaps the most important wild card is the growing rate of inter-ethnic/race marriages, particularly when one parent is non-Hispanic white and the other is not, both because the rates of intermarriage will help determine the future minority composition of the population and because it is unclear how the children of such unions will be classified in future censuses and how they will self-identify as adults.  According to a Pew Research Center report, among all newlyweds in 2010, 9% of whites, 17% of blacks, 26% of Hispanics and 28% of Asians “married out.”  Most of these marriages involved one partner who was non-Hispanic white. The same report notes that during the past 25 years, public sentiment about inter-marriage has changed markedly: in 2010, nearly two-thirds of Americans said it “would be fine” with them if a member of their own family were to marry someone outside their own racial or ethnic group, while in 1986, two thirds of the population held the opposite view.  During the next 25 years, marrying out is likely to become even more common and more widely accepted.

Such trends in inter-marriage tell only part of the story regarding inter-racial childbearing, however. The percentage of births that are non-marital has been increasing steadily since the 1940s, and the greatest increases have taken place in recent years.  Today, 36 percent of all births taking place in the U.S. are non-marital, with much higher percentages among teens (86 percent) and women in their early 20s (62 percent). Given that younger cohorts are more favorably disposed to inter-racial relationships, many of which result in childbearing outside of marriage, the statistics on inter-marriages in the Pew report could well underestimate the implications for future inter-racial/ethnic childbearing.

The number of inter-marriage/partnerships taking place over the next 30 years, the number of children born to these relationships, and how the race/ethnicity of these children get classified in censuses and surveys will fundamentally affect the share of the population that is identified as minority.  More importantly, how these children will self-identify as adults in 2040 is basically unknown.  Census Bureau population projections tacitly assume that this variable is held constant at today’s levels, and that young adults will self-identify in the future the same way that they were identified as children by their parents.

A strong argument has been made that racial and ethnic identity is highly variable over time and depends upon social and political conditions. In 1970, the question on Hispanic origin was added to the Decennial Census, and in 1980 the question on ancestry, both after concerted political lobbying by Latinos in the case of the former and those of European descent for the latter.  The Asian community successfully lobbied to expand the number of Asian options listed in the 1990 census. However, since then the percentage of the population whose ancestry was not identified by the census has increased, slightly between 1980 and 1990, and dramatically between 1990 and 2000 (increasing from 11 percent to 20 percent).  In 2010 the ancestry question was shifted to the American Community Survey.  In that survey ancestry was unidentifiable or not reported for about 12 percent of the population, but only after a persistent effort with follow-up interviews with respondents having not answered this and other questions. Such follow-up was not conducted for the 2000 census.

In short, ancestry appears increasingly to be less important in how Americans identify themselves. And if ancestry, and perhaps by extension race and ethnicity, becomes less important in how we self-identify as Americans, it is entirely likely that in 30 years the percent minority will become a statistic that has become less robust.  Specifically, fewer persons of Hispanic origin might check that box. Likewise, fewer of mixed-race ancestry might identify as such.

A third way that the majority-minority tipping date might be influenced are the apparently arbitrary definitions of whom the Census Bureau counts as minority.  For example, immigrants from Brazil are not counted as minority (Hispanic/Latino) because of Portuguese ancestry, even though most Brazilians also have some indigenous South American native ancestry.  Persons with ancestry in the Middle East and in North Africa are also categorized as white. But a future OMB directive could require that these persons be counted as minorities.

Finally, it has already been demonstrated that such simple factors as questionnaire wording, order of asking questions about nativity, race and ethnicity, and examples used as prompts to questions, will all influence responses.  It is very unlikely that current questionnaire protocols for these items will be used decades in the future.

While there is a lot of uncertainty about the projections 30-50 years into the future, the new projections for the next 10-20 years are likely to be much more accurate.  And they are extremely valuable for showing the magnitude and importance of different immigration assumptions for relatively near-term trends in population growth and for broadly understanding the changing age, race and ethnic composition of the population. Longer-term trends in race and ethnic composition will depend as much on fertility levels, on rates of intermarriage, and on how we think about others and about ourselves, as it does on actual immigration trends.

Leave a comment

Filed under JCHS