by Calculated Risk on 10/18/2018 03:53:00 PM

From housing economist Tom Lawler: Deaths, Immigration, and “the Demographics”

Projections of the US population by characteristics, especially age, are key inputs into intermediate and long term forecasts for such key economic variables as labor force growth, household growth, and social security and Medicare expenditures. For the “adult” population, the key drivers of intermediate term population projections by age are the starting estimates of the population by age, deaths by age, and net international migration by age. Many analysts rely on “official” Census population projections to formulate forecasts of other key economic variables. Unfortunately, such reliance can be problematic, for various reasons. First, Census only releases intermediate and long term population projections every couple of years, and these projections can become quite dated (the projections from the end of 2014, which were the “latest” official projections until March of this year, overstated “actual” population growth from 2014 to 2017 by about one million.) Second, even “recent” official projections can be dated – e.g., the starting point for the Census population projections released earlier this year was the “Vintage 2016” estimates, which have been superseded by the “Vintage 2017” estimates. Third, “official” population projections may have “unrealistic” or out of date assumptions. For example, the latest Census population projections have unrealistic forecasts for US deaths by age, and questionable assumptions about US net international migration by age. And finally, there is considerable uncertainty in the current political environment about the outlook for US immigration policy, and analysts may want to use either their own assumptions about net international migration, or different “scenarios” for net international migration, to generate forecasts of other key economic variables.

While many folks who talk about the “demographic” outlook focus primarily on the current age distribution of the population, net international migration is also a key driver of adult population growth, and labor force growth. For example, the latest US population estimates (July 1, 2017) show that the US “prime” working age population (25-54) increased by 0.35% from July 1, 2016 to July 1, 2017. If there had been no international migration, however, the US prime working age population would actually have declined very slightly over this period.

Given the issues related to official US population projections, analysts attempting to project key economic variables dependent on population projections are faced with challenging choices: either using official projections knowing there are issues with these projections, or producing their own population projections. The latter choice faces its own issues, as analysts would need to formulate their own projections of deaths by age (and whatever other characteristic they are interested in), as well as their own projections of net international migration by age.

Using more realistic assumptions about death rates by age (calibrated to the latest statistics from the National Center for Health Statistics) and what I believe are more realistic assumptions on the age distribution of net international migration (NIM), I have produced US population projections by age under different NIM assumptions through 2021. I have also produced what household growth and labor force growth would be under those scenarios assuming that headship rates and labor force participation rates by age group remained constant at 2017 levels. I did not do this because I am projecting flat headship rates or labor force participation rates, but instead I wanted to show the sensitivity of forecasts to population assumptions.

My starting point was July 1, 2017 (the latest official estimates of the population by age), and I assumed that NIM for 2018 was the same as that shown in the latest Census population projections (since we’re already in the latter part of 2018, sensitivities from 2017 did not seem very useful to me.)

The scenarios I chose were (1) no international migration (this is not necessarily the same as zero net international migration, which only means that immigration inflows equal emigration outflows); (2) NIM equal to those shown in the latest Census projections; and (3) a “Trumpy” scenario, where NIM is lower than the latest Census projections by 20% in 2019, 30% in 2020, and 35% in 2o21.

The results for these various scenarios are shown in the table below. I also including what such “flat headship rates/LFPs” forecasts would have produced for this period if one used either the “official” Census 2014 population projections of the “official” Census 2017 population projections.

Household Growth and Labor Force Growth Rates Projections Assume Flat Headship Rates and LFP Rates by Age, Alternative Scenarios
	Household Growth (number)
	2019	2020	2021	Annual Average
No International Migration, Adjusted Death/NIM	1,010,107	963,727	879,765	951,200
"Trumpy" NIM, Adjusted Death/NIM	1,234,580	1,172,179	1,084,845	1,163,868
Census NIM, Adjusted Death/NIM	1,290,696	1,259,509	1,190,910	1,247,038
Official Census 2017	1,348,905	1,334,801	1,284,162	1,322,623
Official Census 2014	1,460,383	1,446,511	1,393,524	1,433,473

	Labor Force Growth Rate (%)
	2019	2020	2021	CAGR
No International Migration, Adjusted Death/NIM	0.054%	0.037%	0.032%	0.041%
"Trumpy" NIM, Adjusted Death/NIM	0.337%	0.292%	0.275%	0.301%
Census NIM, Adjusted Death/NIM	0.407%	0.400%	0.403%	0.403%
Official Census 2017	0.445%	0.446%	0.458%	0.450%
Official Census 2014	0.530%	0.529%	0.535%	0.531%

Let’s first start with household growth. Analysts using the “official” Census 2014 Population Projections (which were the latest available until this March) could credibly argue that “official” population projections suggested that household formations over the next three years could easily average over 1.4 million.

If those same analysts updated their projections using the Census population projections released this March, they would have reduced their three-year household forecast by about 8%, or about 111 thousand a year.

If those same analysts used NCHS-based death rates by age (instead of the clearly wrong death rates used in the Census 2017 projections), they would have reduced their household growth forecast by an additional 76,000 a year.

If those same analysts believed that a “Trumpy” immigration bill along the lines of certain proposed Republican bills were likely, then they would lower their household growth forecast by ANOTHER 83,000 a year. Now let’s switch to labor force growth. The Census 2014 population projections suggested that even if labor force participation rates by age remained flat at 2017 levels, the labor force would grow by about 0.53% a year over the next three years. That number is reduced to 0.45% under the latest official projections; lowered to 0.40% under a “better assumptions” scenario; reduced to 0.3% under a “Trumpy” scenario; and in the absence of any international migration there would be virtually no labor force growth over the next few years unless labor force participation rates by age increased.

Obviously, the outlook for household growth, labor force growth, and other key economic variable over the next several years is at least partly dependent on projections of the US population. Folks need to realize, however, that the “demographic” outlook is not as clear as some might lead you to believe, and that forecasts based on “official” population projections are not useful just because they are “official.”

I’ll have more on this topic sometime later this year.