Thursday, March 15, 2018

Lawler: "New Long-Term Population Projections Show Slower Growth than Previous Projections but Are Still Too High"

by Bill McBride on 3/15/2018 04:22:00 PM

CR Note: A key takeaway from this excellent analysis by Tom Lawler is that annual household growth will be much lower than previously expected.

From housing economist Tom Lawler: Census: New Long-Term Population Projections Show Slower Growth than Previous Projections but Are Still Too High; Projections Overstate Growth in all Age Groups Save the Very Young and the Very Old

This week the Census Bureau released new projections of the US population, and to the surprise of no one reading this report the new projections show substantially slower population growth than the last set of projections, released at the end of 2014. Virtually all of the slower projected population growth stemmed from a sharp decline in projected net international migration. The new projections show average annual growth in the total US population from 2017 to 2020 (I’m only focusing on the short-term projections) of 2.355 million (compound annual growth rate (CAGR) of 0.72%), down by 271 thousand from the 2.626 million annual projected increase (CAGR of 0.80%) in the 2014 population projections. The latest projection shows average annual Net International Migration (NIM) from July 1, 2017 to July 1, 2020 of 1.006 million, compared to the unrealistically high 1.267 million per year in the 2014 projection. This reduced forecast for NIM reflected recent trends, and did not reflect any possible policy changes.

One of the biggest surprises to folks who follow various demographic data was the projection for deaths in the latest Census report. While data on deaths from the National Center for Health Statistics (NCHS) showed SIZABLE increases in US deaths and death rates over the past few years – with alarmingly increases in death rates for the 15-44 year old age groups – the latest Census projections showed almost no increase in projected US deaths from 7/1/2017 to 7/1/2020 (2.745 million per year in the latest projections vs. 2.721 in the 2014 projections). Provisional data from the NCHS show that US “age-adjusted” death rates continued to increase last year, and these provisional data suggest that US death on the 12 month period ending last June were a staggering 97 thousand higher than those shown in the latest projections.

Before analysts start plugging these new population projections – especially with respect to age – into their various models, they should look at the key assumptions about the components of change that drive these projections, which are deaths, births, and net international migration, to assess whether they seem “reasonable.: These assumptions are available by age and ethnicity in the “datasets” on the Census Population Projections website. With respect to deaths, it is quite clear that the latest Census population projections do not pass the “sniff test” for reasonableness.

Below is a table comparing the Census 2017 Population Projections’ (C2017) assumptions about deaths for selected age groups for the 12-month period ending June 30, 2017 compared to the National Center for Health Statistics tally of US deaths for 2016 (calendar year).

Age GroupCensus 2017 Projections
2016 (Calendar Year)

A few things jump out from this table. First, the Census 2017 Projections (C2017) show a massively larger (and unrealistically high) number of “new-born” infant deaths than that compiled by the NCHS. I pointed this out to Census analysts, and they are aware of this “issue.” (I actually sent the Population Division this table.). Second, the C2017 assumptions show a massively smaller number of deaths for the 15-84 year age groups combined than does the NCHS, with especially large % differences for the 15-44 year age groups. And finally, the C2017 assumptions show significantly more deaths for the very elderly than does the NCHS.

As I showed in an earlier report, death rates as compiled by the NCHS increased sharply in the 15-44 year old age groups over the past several years, and I have no clue why the C2017 projections do not incorporate this increase. Moreover, the C2017 projections assume that death rates for these age groups will decline from the “too low” 2017 levels over the forecast period.

If in fact death rates remain near recent levels – or even if they gradually reverted to 2014 levels – the number of deaths in these age groups would be massively higher than those shown in the C2017 projection, which the number of deaths of the very elderly (85+ years) would be lower (aggregate death would be higher).

What this means, of course, is that if one were to incorporate the higher “actual” death rates the US has recently experienced into population projections over the next several years, the result would be substantially lower projections in the size of the “working age” population and somewhat higher projection for the very elderly and very young.

To give one an idea of how important death rate assumptions are to the near-term population outlook, below is a table of what the C2017 population projections might be if death rates by age were similar to what was experienced. I kept the net immigration assumptions by age from C2017, though are issues with these as well (that’ll be later). I also show the population projections from C2014. (Note: the starting point for C2017 was “Vintage 2016,” and population estimates were revised upward slightly in “Vintage 2017,” which will also be subject to revision next year. Also, note population numbers are as of July 1.)

Census 2014 Population Projections (000s)

Census 2017 Population Projections (000s)

Updated Projections Assuming 2016 NCHS Death Rates, C2017 NIM (000s)

Census’ 2014 Projections, total population from 2016 to 2020 was forecast to rise by 10.508 million, compared to 9.427 million in the latest projection and 8.962 million if the latest projection held 2016 NCHS death rates constant. The so-called “prime-age” working population was project to rise by 2.053 million in C2014 Projections, compared to 1.524 million in the C2017 Projection and 1.303 million if the C2017 has assumed constant 2016 death rates.

Annual Growth Rate, 25-54 Year Old Population
C2017 *
CAGR, 2016-20200.40%0.30%0.25%
*Assumes 2016 NCHS Death Rates

Obviously, an updated population projection from those from 2014 produces slower projections for labor force growth, and significantly slower if one uses “realistic” growth rates. Similarly, household projections using 2014 projections are a LOT higher than those using updated assumptions and realistic death rates. E.g., from mid-2018 to mid-2020 a reasonable projection for annual household growth using the 2014 population projections of about 1.455 million. Using the “raw” 2017 projections would, using similar headship rates, produce an annual household growth forecast of about 1.345 million. Adjusting the 2017 projections for more realistic death rates, however, would (using same headship rates) result in an annual household growth forecast over the next two years of about 1.245 million, or about 210,000 a year less than one would get using the old C2014 projections.

I’ll have more on the topic of population projections later.