by Calculated Risk on 8/23/2019 04:00:00 PM

From housing economist Tom Lawler: Updated “Demographic” Outlook Using Recent Population Estimates by Age

Executive Summary: Analysts who use intermediate or long term population projections to forecast key economic variables such as labor force growth, household growth, etc. should recognize that the latest official Census intermediate and long term population projections (produced in 2017 and referred to as “Census 2017”) are out of date. Specifically, Census 2017 materially over-predicted births, materially under-predicted deaths (mainly for non-elderly adults), and somewhat over-predicted net international migration (NIM) for each of the last several years. In addition, the assumptions in Census 2017 projections over the next several years (and more) are almost certainly too high for births, too low for deaths, and too high for NIM. As a result, population growth, household growth, and labor force growth over the next few years will be lower than forecasts based on the Census 2017 population projections. How much lower depends critically on net international migration, which in the current environment is a big unknown.

Using more realistic assumptions on births and deaths by age, I have developed updated population projections by age through 2021 assuming (1) net international migration in each year is the same as in 2018; and (2) there is no international migration in 2020 or 2021. I did the latter scenario to highlight the importance of net international migration assumptions on population projections.

--------------------------------------------------------------------------------------------

Earlier this year the Census Bureau released its latest (“Vintage 2018) estimates of the US resident population by single year of age for July 1, 2018, as well as for July of each of the previous 8 years. These latest estimates give analysts a new starting point that can be used to update population projections by age using assumptions about births, deaths by age, and net international migration by age. These population projections are key inputs into forecasts of other key economic variables such as the labor force and US households.

While many analysts prefer to use “official” Census population projections in forecasting other key economic variables, there are several reasons why this is often not a good idea. First, official Census population projections are only released every couple of years, and may be out of date. And second, such projections may have assumptions about the key drivers of population growth that may not be viewed as “reasonable.”

The latest official Census population projections were done in late 2017 and released to the public in early 2018. The “starting point” for these projections was the “Vintage 2016” population estimates, and population estimates for 2016 have since been revised. In addition, current estimates of births, deaths, and net international migration from 2016 to 2018 are significantly different from the “Census 2017” projections. And finally, the assumptions in the “Census 2017” projections for the key drivers of population changes, especially death rates by age, are not realistic or consistent with recent actual death rates by age.

The latest estimate of the US resident population on July 1, 2018 was 327,167,434, which is 724,477 lower than the Census 2017 projection for that date. Fewer births, more deaths, and lower international migration all contributed to the projection shortfall. Below is a table showing the differences between the latest July 1 2018 population estimate and the July 1, 2018 forecast from the Census 2017 projection by key component.

	Census 2017 Projections	Vintage 2018 Estimates	Difference
7/1/2016	323,127,513	323,071,342	(56,171)
Births: 7/1/2016 - 6/30/2018	8,129,169	7,757,482	(371,687)
Deaths: 7/1/2016 - 6/30/2018	5,363,099	5,593,449	230,350
Net Int'l Migration: 7/1/2016 - 6/30/2018	1,998,328	1,932,059	(66,269)
7/1/2018	327,891,911	327,167,434	(724,477)

Births from 7/1/2016 to 6/30/2018 were significantly below projected levels from Census 2017. In addition, the National Center for Health Statistics (NCHS) recently estimated that US births in 2018 (calendar year) totaled just 3,788,235, the lowest annual number of births in 32 years, and a whopping 306,730 below the Census 2017 projection for the 12 month period ending June 2019.

Deaths from 7/1/2016 to 6/30/2018 were significantly above projected levels from Census 2017. While the Census Bureau did not release estimates of deaths (or net international migration) by age in its “Vintage 2018” release, Census does use data from the NCHS on deaths by age, and these data indicate that most of the higher than projected deaths from Census 2017 were in the 20-74 year old age groups.

The Census 2017 population projections were based on a dated “death rate” table, as well as on projections that death rates for most age groups would decline each year. In fact, however, death rates for many age groups have increased over the past few years. The assumptions on deaths from Census 2017 over the next several years are almost certainly way too low, especially for the 20-74 year old age groups.

Finally the latest Census estimates of net international migration (NIM) from 7/1/2016 to 6/30/2018 are somewhat below the Census 2017 projections. Moreover, updated estimates for NIM (which are, unfortunately, subject to considerable error) suggest a materially different age distribution than that assumed in the Census 2017 projections.

Below is a table comparing the latest Census estimates of the US resident population for 2018 with the projections from Census 2017 for 2018 by 5-year ago groups.

US Resident Population, 7/1/2018
	Vintage 2018 Estimate	Census 2017 Projection	Difference
Total	327,167,434	327,891,911	-724,477
0-4	19,810,275	20,172,617	-362,342
5-9	20,195,642	20,166,270	29,372
10-14	20,879,527	20,866,300	13,227
15-19	21,097,221	21,084,451	12,770
20-24	21,873,579	21,966,919	-93,340
25-29	23,561,756	23,601,976	-40,220
30-34	22,136,018	22,125,395	10,623
35-39	21,563,587	21,556,772	6,815
40-44	19,714,301	19,728,816	-14,515
45-49	20,747,135	20,786,395	-39,260
50-54	20,884,564	20,923,227	-38,663
55-59	21,940,985	22,012,901	-71,916
60-64	20,331,651	20,408,221	-76,570
65-69	17,086,893	17,127,778	-40,885
70-74	13,405,423	13,418,850	-13,427
75-79	9,267,066	9,274,592	-7,526
80-84	6,127,308	6,131,125	-3,817
85+	6,544,503	6,539,306	5,197

As the table shows, there are significant differences between the “Vintage 2018” population estimates and the projections from Census 2017, not just in the total but also in the age distribution.

In “Vintage 2018” Census also provided updated population projections for 2019, which are used (among other things) as “controls” for the household employment estimates for 2019. Below is a table comparing the Vintage 2018 population projections for 2019 with the Census 2017 projections for 2019.

Resident Population Projections for 7/1/2019
	Vintage 2018	Census 2017	Difference
Total	329,158,518	330,268,840	-1,110,322
0-4	19,702,853	20,304,120	-601,267
5-9	20,222,613	20,180,503	42,110
10-14	20,829,999	20,814,198	15,801
15-19	21,114,493	21,100,800	13,693
20-24	21,744,207	21,861,558	-117,351
25-29	23,612,082	23,693,302	-81,220
30-34	22,510,095	22,522,671	-12,576
35-39	21,798,813	21,798,185	628
40-44	19,979,623	20,003,487	-23,864
45-49	20,443,042	20,491,118	-48,076
50-54	20,515,438	20,559,577	-44,139
55-59	21,912,472	22,001,226	-88,754
60-64	20,608,057	20,712,610	-104,553
65-69	17,487,028	17,543,604	-56,576
70-74	14,060,332	14,076,777	-16,445
75-79	9,678,247	9,685,598	-7,351
80-84	6,327,181	6,329,092	-1,911
85+	6,611,943	6,590,414	21,529

(Note: The Vintage 2018 projection for 2019 appears to have assumed the same number as births as for 2018, though recent data suggest that births were lower.)

If analysts had used the Census 2017 population projections to forecast the US labor force and the number of US households, and had been accurate in their forecasts of labor force participation rates and headship rates, they would have over-predicted the size of the labor force in mid-2019 by about 400,000, and over-predicted the number of households in mid-2019 by about 260,000.

Obviously, Census 2017 population projections have not tracked recent estimates and projections very well. In addition, Census 2017 assumptions for births, deaths, and net international migration are likely to be considerable off from likely “actuals” for the years ahead.

For analysts who use intermediate and long term population projections to forecast other key economic or social variables such as household growth, labor force growth, social security/medicare enrollment/payments, college enrollment, etc., it seems clear that it would not be appropriate to use the Census 2017 population projections. However, these are the latest “official” projections that have been released, and many analysts prefer to use “official” projections. Moreover, formulating one’s own population projections by age requires one to make assumptions not just on total births, deaths, and NIM, but also deaths and NIM by single year of age, and there aren’t timely publicly-released data on either of the latter.

To help some of these analysts, I have, using some unpublished data on recent trends, produced US resident population projections through 2021 using the following assumptions:
1. Annual births from 2019 through 2021 are the same as those in calendar year 2018 (3,788,235);
2. Deaths rates by age are similar to those in 2018 (though somewhat lower for age groups that saw a sizable increase over the last few years); and
3. Net International Migration by age is the same each year as the Census estimates for 2018.

Note that the biggest “wild card” in assumptions is NIM; not only are recent estimates subject to much higher uncertainty than the other two key drivers of population growth, but it is also virtually impossible in the current political environment to make a reasonable projection for NIM. For example, recent actions by the Administration set to take effect in mid-October would, if implemented, have a significantly negative impact on immigration over the next few years.

I also have produced population projections by age assuming no international migration (this is not necessarily the same as no immigration, as a lot of people leave the country for abroad each year). I did this to highlight the importance of NIM on the outlook for the population.

Note that I did not use the Vintage 2018 projections for 2019, but instead used the assumptions discussed above.

These alternative projections are shown on the next page for select age groups.

US Resident Population: Alternative Projections
Census 2017 Projections
	7/1/2018	7/1/2019	7/1/2020	7/1/2021
Total	327,891,911	330,268,840	332,639,102	334,998,398
0-14	61,205,187	61,298,821	61,408,926	61,510,603
15-24	43,051,370	42,962,358	42,937,831	43,004,867
25-34	45,727,371	46,215,973	46,491,403	46,716,390
35-44	41,285,588	41,801,672	42,351,795	43,006,437
45-54	41,709,622	41,050,695	40,615,037	40,324,022
55-64	42,421,122	42,713,836	42,782,544	42,593,657
65-74	30,546,628	31,620,381	32,789,437	33,953,050
75+	21,945,023	22,605,104	23,262,129	23,889,372
Flat Births, More Realistic Death Rates, Flat NIM
	7/1/2018	7/1/2019	7/1/2020	7/1/2021
Total	327,167,434	329,072,705	330,917,403	332,700,753
0-14	60,885,444	60,698,571	60,507,167	60,290,588
15-24	42,970,800	42,861,557	42,822,738	42,882,730
25-34	45,697,774	46,128,836	46,337,496	46,484,660
35-44	41,277,888	41,786,370	42,323,937	42,961,919
45-54	41,631,699	40,958,397	40,507,001	40,198,337
55-64	42,272,636	42,516,252	42,534,068	42,293,993
65-74	30,492,316	31,537,868	32,667,883	33,780,930
75+	21,938,877	22,584,854	23,217,113	23,807,596
Flat Births, More Realistic Death Rates, No International Migration
	7/1/2018	7/1/2019	7/1/2020	7/1/2021
Total	327,167,434	329,072,705	329,940,031	330,747,524
0-14	60,885,444	60,698,571	60,284,355	59,860,569
15-24	42,970,800	42,861,557	42,537,458	42,330,317
25-34	45,697,774	46,128,836	46,069,155	45,934,117
35-44	41,277,888	41,786,370	42,209,180	42,720,399
45-54	41,631,699	40,958,397	40,468,714	40,117,461
55-64	42,272,636	42,516,252	42,507,588	42,241,276
65-74	30,492,316	31,537,868	32,649,225	33,742,244
75+	21,938,877	22,584,854	23,214,356	23,801,141

Below are tables showing the differences between the latter two scenarios and the Census 2017 population projections.

Alternate Population Projections vs. Census 2017 Projections
Flat Births, More Realistic Death Rates, Flat NIM
Total	-724,477	-1,196,135	-1,721,699	-2,297,645
0-14	-319,743	-600,250	-901,759	-1,220,015
15-24	-80,570	-100,801	-115,093	-122,137
25-34	-29,597	-87,137	-153,907	-231,730
35-44	-7,700	-15,302	-27,858	-44,518
45-54	-77,923	-92,298	-108,036	-125,685
55-64	-148,486	-197,584	-248,476	-299,664
65-74	-54,312	-82,513	-121,554	-172,120
75+	-6,146	-20,250	-45,016	-81,776
Flat Births, More Realistic Death Rates, No International Migration
	7/1/2018	7/1/2019	7/1/2020	7/1/2021
Total	-724,477	-1,196,135	-2,699,071	-4,250,874
0-14	-319,743	-600,250	-1,124,571	-1,650,034
15-24	-80,570	-100,801	-400,373	-674,550
25-34	-29,597	-87,137	-422,248	-782,273
35-44	-7,700	-15,302	-142,615	-286,038
45-54	-77,923	-92,298	-146,323	-206,561
55-64	-148,486	-197,584	-274,956	-352,381
65-74	-54,312	-82,513	-140,212	-210,806
75+	-6,146	-20,250	-47,773	-88,231

Obviously, the outlook for population growth, labor force growth, household formations, and other economic variables over the next few years depends critically on one’s assumptions about net international migration. The “flat births/flat NIM” scenario is probably a “high” forecast, given, recent Trump administration actions/policies, while the “no international migration” scenario is more designed to show what population growth would look like without international migration. In the “flat births/flat NIM” scenarios growth in the labor force over the next two years would be about 0.1% lower per year than forecasts based on Census 2017, while household growth would be about 120,000 lower per year. In the no international migration scenario labor force growth over the next two years would be 0.4% lower per year, and household growth would be about 370,000 per year lower per year, than Census 2017-based forecasts.