by Calculated Risk on 3/06/2018 02:03:00 PM

From housing economist Tom Lawler: Population Outlook: Uncertainty Not Just Related to Immigration Assumptions, But Also Death Rates

In its latest “Tip Sheet” Census said that it will release new long-term population projections sometime this month. The last official Census long-term population projections were released at the end of 2014, and since then “actual” population estimates have fallen way short of those projections. As I documented in a previous report, the population shortfall from that projection was the result of lower births, higher deaths, and (especially) lower net international migration. I also noted that updated net international migration projections going forward are likely to be massively lower than those shown in the 2014 projections, and that analysts should not just look at the overall population projections but also the net international migration assumptions in assessing the “reasonableness” of the projections.

Another factor that will almost certainly result in lower population projections relative to those from 2014 will be deaths. The 2014 Census population projections assumed that the “death rates” for most age groups, and especially the young to middle age groups, would gradually decline over time. In fact, however, death rates for young and middle age adults have increased considerably since 2014, for some age groups at an alarming rate. As a result, despite lower overall population counts from lower net international migration, the number of deaths has considerably exceeded the Census 2014 projections.

Annual (calendar year) data on US deaths are compiled by the National Center for Health Statistics (NCHS), though data are only available with a considerable lag. For example, the NCHS released a “data brief” on mortality in the US for 2016 last December, and detailed data on US deaths for 2015 were released last year. The NCHS data are used by the Census Bureau to estimate and project the overall US population and the characteristics of the population, though Census must make assumptions for the most recent years. Also, Census death estimates are for the 12-month period ending on June 30th, as opposed to the NCHS calendar-year estimates. Below is a table showing annual deaths from the NCHS; death assumptions from the Census 2014 population projections; and the death estimates from the latest Census population estimates (Vintage 2017).

12-month period ending:	NCHS 31-Dec	CB Vintage 2017 30-Jun	CB 2014 Projection 30-Jun
2014	2,626,418	2,582,448	2,582,858
2015	2,712,630	2,699,826	2,619,062
2016	2,744,248	2,728,714	2,649,979
2017		2,744,040	2,680,555

As the table indicates, Census projections for US deaths have been well south of “actual” deaths. While Census does not release its assumptions about deaths by age group (or other characteristics) in its current “estimates” report, it does show its assumptions for deaths by age in its periodic long-term population projections. The NCHS also releases data on deaths by age groups. Below is a comparison of NCHS deaths by selected age groups with the death assumptions from the Census 2014 population projections.

NCHS Data on Deaths by Selected Age Groups (Calendar Year)
	2014	2015	2016	2016 vs. 2014
Total	2,626,255	2,712,630	2,744,248	117,993
15-24	28,791	30,494	32,575	3,784
25-34	47,177	51,517	57,616	10,439
35-44	70,996	73,088	77,792	6,796
45-54	175,917	174,494	173,516	-2,401
55-64	348,808	357,785	366,445	17,637
65-74	471,541	495,016	512,080	40,539
75-84	624,504	637,566	636,916	12,412
85+	826,226	859,701	854,462	28,236

Census 2014 Projections of Deaths by Selected Age Group (12-month period ending 6/30)
	2014	2015	2016	2016 vs. 2014
Total	2,582,858	2,619,062	2,649,979	67,121
15-24	27,542	26,966	26,150	-1,392
25-34	43,783	43,490	43,107	-676
35-44	66,534	64,991	63,159	-3,375
45-54	170,644	165,999	160,879	-9,765
55-64	326,669	328,230	328,830	2,161
65-74	453,167	469,548	485,155	31,988
75-84	621,098	625,385	630,841	9,743
85+	840,781	861,957	879,520	38,739

As the above table show, the Census 2014 population projections assumed that deaths of people aged 15-44 years old would decline from 2014 to 2016 – reflecting an assumption of declining death rates in those age groups – while NCHS data indicate that deaths in these age groups rose significantly over this period, reflecting sizable increases in death rates. On the next page is a table showing some historical NCHS data on death rates by selected age groups. Note that these death rates are from past reports using available population estimates at the time of the report, and that death rates using revised population estimates would be slightly different for past years. Note also that for 2016 NCHS has only released death rates for newborns and for 10-year age groups, and I have “guesstimated” 5-year age group death rates for 2016 based on the 10-year death rate estimates.

Deaths per 100,000 population, Selected 5-year Age Groups, NCHS
Age Group	2010	2011	2012	2013	2014	2015	2016 (est)
15-19	49.4	48.9	47.2	44.8	45.5	48.3	49.0
20-24	86.5	86.1	84.6	83.4	83.8	89.3	90.3
25-29	96.0	96.7	98.1	97.6	99.7	106.4	110.2
30-34	110.2	113.0	112.9	114.8	117.3	127.4	129.7
35-39	138.8	140.2	140.5	141.9	147.2	154.2	157.1
40-44	201.1	201.7	198.7	200.3	202.4	206.1	216.0
45-49	324.0	323.5	315.7	315.1	311.3	308.7	312.5
50-54	491.7	494.6	491.5	491.6	491.3	493.0	493.1
55-59	711.7	712.3	717.2	721.8	730.6	731.8	737.7
60-64	1015.8	1005.5	1014.0	1021.7	1032.2	1039.3	1042.2
65-69	1527.6	1502.9	1470.2	1468.4	1454.0	1465.6	1447.4
70-74	2340.9	2306.2	2266.7	2261.7	2246.1	2260.1	2235.8
75-79	3735.4	3685.0	3630.9	3611.9	3560.5	3574.1	3479.3
80-84	6134.1	6116.8	6026.1	6027.4	5944.6	5986.2	5809.1
85+	13934.3	13779.3	13678.6	13660.4	13407.9	13637.9	13392.1

As the table indicates, death rates for all 5-year age groups under 45 years old increased significantly from 2014 to 2016 (especially for 25-34 year olds), while death rates for the elderly actually declined somewhat. This contrasts sharply with the CB 2014 population projections, which assumed that death rates would decline for all age groups save the very elderly.

While it is beyond the scope of this report to discuss why death rates among all but the very old have risen sharply rather than decline as assumed in the CB 2014 population projections (though drug overdose deaths are a significant component – see below), it seems almost certain that the updated population projections will incorporate substantially higher death rates for most of the population – and especially for the “prime” working age population. And, as discussed in an earlier report, the updated population projections will almost certainly assume much lower net international migration assumptions than was the case in the 2014 population projections. This, too, will have a “disproportionate” impact on projections for “young” adults and the “prime” working age population.

Drug Overdose Deaths by Selected Age Groups, NCHS
	15-24	25-34	35-44	45-54	55-64	65+	Total
2012	3,518	8,508	8,948	11,895	6,423	2,094	41,386
2013	3,664	8,947	9,320	12,045	7,551	2,344	43,871
2014	3,798	10,055	10,134	12,263	8,122	2,568	46,940
2015	4,235	11,880	11,505	12,974	8,901	2,760	52,255
2016	5,376	15,443	14,183	14,771	10,632	3,075	63,480