Wednesday, February 08, 2017

Lawler on Household Projections

by Bill McBride on 2/08/2017 10:51:00 AM

From housing economist Tom Lawler: Household Projections: New Population Estimates + New Administration = Time for an Update

When the Census Bureau released its estimates for the U.S. population in December, most press coverage focused on which states saw the fastest population growth last year. What many missed, however, was that the Census Bureau significantly reduced its population estimates for each of the past several years, with the major reason for the downward revisions stemming from reduced estimates of net international migration. The latter reductions were the result up an updated methodology used to estimate foreign-born emigration, as discussed in the following excerpt from the 2016 vintage “release notes.”

“The Vintage 2016 net international migration estimates reflect the following changes to the methodology since the release of the Vintage 2015 estimates:

“We updated the foreign-born emigration subcomponent in two ways: 1) we modified the emigrant group definitions used to calculate estimates of foreign-born emigration; 2) we applied averaged rates from multiple 5-year ACS files for non-recent arrivals (Mexican born who arrived more than 10 years ago, Asian born who arrived more than five years ago, and Non-Mexican born who arrived more than 10 years ago). These changes resolve negative rates produced by the previous residual method, which had resulted in zero emigration for certain emigrant groups. Consequently, foreign-born emigration will be higher and net international migration will be lower than the previous vintage.”
Here are some summary statistics on “Vintage 2016” population estimates (resident population” compared to “Vintage 2015” population estimates. I’m including Census projections for 2016 that had been based on Vintage 2015 estimates (used for CPS-based data for 2016) as well as a December 1, 2016 estimate based on Vintage 2016.

U.S. Resident Population, Vintage 2015 vs. Vintage 2016
  Vintage 2015Vintage 2016Change
7/1/2010309,346,863309,348,1931,330
7/1/2011311,718,857311,663,358-55,499
7/1/2012314,102,623313,998,379-104,244
7/1/2013316,427,395316,204,908-222,487
7/1/2014318,907,401318,563,456-343,945
7/1/2015321,418,820320,896,618-522,202
7/1/2016323,889,854323,127,513-762,341
12/1/2016325,032,763324,142,480-890,283

The next table shows Census estimates of Net International Migration from the “Vintage 2015” population estimates compared to the most recent (“Vintage 2016”) estimates.

Net International Migration, July 1 - July 1, Vintage 2015 vs. Vintage 2016
  Vintage 2015Vintage 2016Change
2011910,951854,172-56,779
2012948,321899,576-48,745
2013992,215873,972-118,243
20141,133,261977,801-155,460
20151,150,5281,036,826-113,702
2016*1,160,000999,163-160,837
*Estimate for Vintage 2015 projections

These downward population revisions were reflected in last Friday’s employment report, which showed the impact of updated population estimates on the 16+ year civilian non-institutional population assumptions used to produce employment and labor force estimates based on the “household” (CPS) survey.

16+ Civilian Non-Institutional Population, Household Employment Report, December 2016 (000’s)
  Vintage 2015Vintage 2016Change
Total254,742253,911-831
White198,845198,376-469
Black32,10532,029-76
Asian15,43315,175-258
Other8,3598,331-28
Hispanic or Latino Ethnicity*41,19040,838-352
*Persons whose ethnicity is described as Hispanic or Latino may be of any race

While Census has not yet updated its estimates of age distribution of the population, data released as part of last Friday’s report allow one to estimate downward revisions in the 16+ civilian non-institutional population by age.

16+ Civilian Non-Institutional Population by Age Group Derived from* Household Employment Report, December 2016 (000's)
Age GroupVintage 2015Vintage 2017Change
16-2438,34838,265-83
25-3443,83943,692-147
35-4439,91939,763-156
45-5442,21242,103-109
55-6441,58741,468-119
65-7428,91328,794-119
75+19,92419,826-98
Total254,742253,911-831
*LEHC estimates

These new population estimates will, of course, be used by Census to produce household estimates based on the 2017 CPS/ASEC. However, Census typically does not revise previous-year estimates to reflect revisions in population estimates. As such, the change in the CPS/ASEC household estimate for March 2017 compared to the March 2017 estimate will probably be about 450,000 or so lower than it would have been without the recent methodological change used to estimate net international migration

These latest population revisions, including updated projections from Census for 2017, suggest that household projections based on the latest long-term population projections made by Census in December 2014 are woefully out of date. Here is a table comparing the population projections in that report for 2014, 2015, 2016, and 2017 compared to the most recent population estimates and projections (the latter of which only extend to 2017).

2015 Population Projections vs. Latest Estimates/Projections, U.S. Resident Population (000’s)
  Projection from
December 2014
Latest Estimates/
Projections
Change
7/1/2014318,748318,563-185
7/1/2015321,369320,877-492
7/1/2016323,996323,128-868
7/1/2017326,626325,534-1,092

As the table shows, the most recent projection of the US resident population for 2017 is almost 1.1 million lower than the projection from late 2014.

Census Population ChangeGiven (1) the improved methodology to estimate net international migration, and (2) the Trump administration’s potential policies on immigration, it seems extremely likely that an updated long-term population projection would produce hugely different population projections for years subsequent to 2017 than those shown in the “latest” Census projections from late 2014. E.g., here is a table showing the components of change from Census’ long-term population projections from late 2014.

While updated estimates suggest that net international migration averaged about 900,000 over the past 5 years, the late 2014 projections assumed an average of about 1.256 million from 2015 to 2020. Even without the “Trump win” that number would currently be considered way to high, and today such a projection seems, in the words of demographer Dr. Vizzini, “inconceivable.”

For folks wondering why I am comparing recent estimates to a Census projection from December 2014, the reason is that the December 2014 projection is that latest the Census has released, and those projections have been used by quite a few analysts/institutions/organizations/etc. to produce and publish long-term projections of the number of US households. Indeed, in an unluckily-timed report released just eight days before the Census released it “Vintage 2016” population numbers, the Joint Center for Housing Studies published new household projections based on Census population projections from December 2014! (There are other serious problems with the JCHS report, but I won’t dwell on those).

So ... If one both incorporated the latest Census population projections (including its new methodology for measuring net international migration), AND used updated assumptions on births, deaths, and net international migration for the next several years, what would an updated population projection look like? E.g., how should one translate what Trump and Trump officials (as well as Congressional officials) have said about immigration into a projection of net international migration? Frankly, I don’t know.

But let’s just make an “educated guess” that net international migration would be at a level lower than it recently has been, and let’s arbitrarily pick an average of 750,000. Combed with reasonable projections for birth and dealt rates, that would produce a population projection of something like that shown in the table below. (I am not adjusting Census’ latest projection for 2017).

US Resident Population Estimates and Projections (000's)
December 2014
Projection
"More Trumpian"Difference
7/1/2015321,369320,877-492
7/1/2016323,996323,128-868
7/1/2017326,626325,534-1,092
7/1/2018329,256327,647-1,609
7/1/2019331,884329,751-2,133
7/1/2020334,503331,840-2,663

If such a lower net international migration number were to occur, then a “more reasonable” projection of the US resident population in 2020 would be a whopping 2.663 million below the projection in Census’ December 2014 population forecast used by quite a few analysts to project US household growth. The vast bulk of this reduction would be (1) in the adult population, and (2) in the foreign-born population. I have not yet attempted to see how such a different population projection would translate into a household projection, but some might call it “YUGE” enough to warrant throwing out household projections based on the “latest official” Census population projections, and instead focus on an updated forecast based on more reasonable population projections.