by Calculated Risk on 6/16/2017 03:48:00 PM

From housing economist Tom Lawler: Reasonable Population Projections Are Important!

One of the key variables analysts look at in attempting to project such things as housing demand, labor force growth, government “entitlement” spending, and a host of other key “macro” indicators is the projected population -- not just (or even mainly) the total population, of course, but more especially the age distribution of the population. At first glance, one might think that projecting the age distribution of the population would be rather simple: if you know your country’s current age distribution, you should, if your country has “closed borders,” be able to “walk forward” each individual age “cohort” using reasonable assumptions of death rates by age. However, we are not a “closed border” country, and as such, projections of the population by age also need to take into account “net international migration” by age, which encompasses both immigration by age and emigration (folks moving out of the country) by age.

Rather than making their own assumptions about death rates and net international migration, many analysts rely on “official” Census long-term population projections to produce forecasts of other variables. Sadly, however, such “official” projections are done infrequently, and the last “official” Census population projection was from late 2014, and is extremely out of date.

To remind folks, the 2014 Census long-term population projection incorporated an assumption that net international migration would increase materially beginning in 2014, and gradually increase from these unusually high levels through the next several decades. Updated population estimates through 2016, however, suggest that these net international migration assumptions through 2016 were way too high. In addition, recent data suggest that death rates over the past two years have been higher than those incorporated in the Census 2014 projections. As a result, the latest estimate of the US population as of July 1, 2016 is 868,105 lower than the projection from 2014, with the bulk of this projection “miss” coming in the 20-64 year old range.

In addition, the Census 2014 population projections did not (for obvious reasons) incorporate an assumption that Donald Trump would be elected president and that the House and Senate would be controlled by Republicans, and what that combination might mean in terms of likely net international migration trends over the remainder of the decade. Suffice it to say, the net international migration assumptions from the Census 2014 population projections from 2016 on are unrealistically high.

In sum, as is obvious, the “latest” long-term population projections from Census are extremely out of date, and are of little if any use to anyone. While Census does plan to issue a new long-term population projection around the end of this year, in the interim any analysts who wish to issue projections of any variables that depend on population forecasts must produce their own population projections.

Click on graph for larger image.

Unfortunately for analysts, there is no clear or logical methodology to project net international migration over the next few years based on either the rhetoric or the policy “statements” from the Trump administration. However, one thing analysts can do is to show different population projection “scenarios” based on different net international migration assumptions.

One must first, of course, start with the latest available estimates of the US population, which are for July 1, 2016, and which are shown in the chart below.

The chart highlights two “demographic” facts most folks are aware of: the high number of so-called “baby-boomers” set to near of reach retirement over the next several years, and the large number of “echo-boomers” who are currently “young adults.”

For sensitivity analysis, however, there are “shortcomings” of doing various scenarios from July 1, 2016: first, of course, that date is now almost a full year ago; and second, Trump “policies” probably have had only a modest impact on net international migration from July 1, 2016 until now.

As such, the “scenarios” I am showing are from July 1, 2017 to July 1, 2020. In estimating the age distribution of the population for 2017 I am assuming net international migration of 850,000, which is about 150,000 less than the latest estimate for 2016. I am using a similar age distribution for net international migration as that used in the Census 2014 projection, but I have adjusted assumptions on death rates to be consistent with the latest estimate for deaths in 2016 (which was higher than that assumed in the Census 2014 projection).

The two scenarios I am showing for the US resident population by selected age groups are (1) zero net international migration (which in essence show how the existing population “ages); and (2) net immigration per year of 700,000 from July 1, 2017 through July 1, 2020. This assumption is, admittedly, somewhat arbitrary, but is “consistent” with a view that net international migration will be lower over the next 3 years than any three year period over the past two decade. For comparison I am also showing the population projections for 2016 through 2020 from the Census 2014 long-term population projection release.

US Resident Population Projection by Selected Age Groups, Various "Scenarios", Thousands
Zero Net International Migration, 7/1/2017 - 7/1/2017
	07/01/16	07/01/17	07/01/18	07/01/19	07/01/20	7/1/2017 - 7/1/2010
Total	323,128	325,148	326,287	327,387	328,487	3,339
0-14	60,975	60,954	60,792	60,612	60,492	-463
15-24	43,511	43,149	42,660	42,274	41,977	-1,172
25-34	44,677	45,185	45,362	45,530	45,463	278
35-44	40,470	40,760	41,138	41,514	41,909	1,149
45-54	42,787	42,323	41,638	40,907	40,395	-1,927
55-64	41,463	41,986	42,337	42,556	42,542	556
65-74	28,630	29,667	30,497	31,530	32,649	2,982
75+	20,614	21,124	21,863	22,464	23,060	1,936
25-54	127,934	128,268	128,138	127,951	127,767	-500
Net International Migration of 700,000 per Year from 7/1/2017 to 7/1/2020
	07/01/16	07/01/17	07/01/18	07/01/19	07/01/20	7/1/2017 - 7/1/2010
Total	323,128	325,148	326,987	328,785	330,580	5,433
0-14	60,975	60,954	60,906	60,832	60,811	-143
15-24	43,511	43,149	42,863	42,663	42,532	-618
25-34	44,677	45,185	45,544	45,908	46,048	863
35-44	40,470	40,760	41,223	41,690	42,183	1,423
45-54	42,787	42,323	41,682	40,997	40,533	-1,789
55-64	41,463	41,986	42,373	42,626	42,648	662
65-74	28,630	29,667	30,522	31,583	32,730	3,063
75+	20,614	21,124	21,874	22,486	23,095	1,971
25-54	127,934	128,268	128,449	128,595	128,764	497
Projections from Census 2014 Long-Term Population Forecast
	07/01/16	07/01/17	07/01/18	07/01/19	07/01/20	7/1/2017 - 7/1/2010
Total	323,996	326,626	329,256	331,884	334,503	7,878
0-14	61,037	61,176	61,314	61,435	61,577	401
15-24	43,613	43,352	43,202	43,125	43,107	-245
25-34	44,865	45,490	46,018	46,561	46,890	1,400
35-44	40,578	40,930	41,477	42,035	42,628	1,698
45-54	42,864	42,443	41,860	41,239	40,842	-1,601
55-64	41,619	42,180	42,619	42,931	43,019	840
65-74	28,747	29,825	30,743	31,860	33,075	3,250
75+	20,674	21,230	22,023	22,697	23,366	2,136
25-54	128,306	128,863	129,356	129,836	130,360	1,496

To give folks an idea of how much different, below is a table of what labor force growth and US household formations would be under each scenario if (1) labor force participation rates by age remained constant at 2016 levels; and (2) household headship rates by age remain constant at my “best guess” rates for 2016 (there are no good, reliable data on households since 2010, but that is a different story!). I realize, of course, that holding labor force participation rates and headship rates by age constant is not a “best guess” projection, but I’m just trying to show sensitivities to different population assumptions.

Annual Growth Rate in the US Labor Force Assuming Constant Labor Force Participation Rates by Age
	2018	2019	2020
Zero Net International Migration	0.04%	0.02%	-0.01%
NIM of 700,000/year	0.28%	0.27%	0.24%
Census 2014 Projections	0.50%	0.50%	0.48%
US Household Growth Assuming Constant Headship Rates by Age
	2018	2019	2020
Zero Net International Migration	1,026,077	966,155	924,937
NIM of 700,000/year	1,231,995	1,180,916	1,148,645
Census 2014 Projections	1,485,278	1,455,615	1,442,362

As the table suggests, analysts using the extremely dated Census 2014 population projections would conclude that the US would have “decent” labor force growth and quite strong US household growth over the next three years. Contrary to what some analysts suggest, however, that strong growth is not in the main the result of the current “demographics” of the population, but rather is mainly the result of what are now clearly unrealistically high assumptions about net international migration. If instead the US had zero net international migration of the next three years, the US labor force would show no growth unless labor force participation rates increased, and US household growth would average less than one million per year unless headship rates increased. Not surprisingly, a “sorta Trumpy” scenario of net international migration of 700,000 a year – probably the closest there is a a “base case” scenario” produces projections about half way in between these two extremes.

Folks may differ about what is or is not a “reasonable” assumption for net international migration over the next three years, but what seems almost crystal clear is that the assumptions in the latest official long-term population projections from Census released in 2014 are not realistic, and should not be used by competent analysts. That is why it was so surprisingly that in today’s release of its annual “The State of the Nation’s Housing,” the Joint Center for Housing Studies at Harvard University stuck with the housing projections it made near the end of last year, which used … you guessed it --- the Census 2014 long-term population projections! While the JCHS report did note that net international migration could well be a LOT lower than that assumed in the Census 2014 population projections, its report did not give any indication of what a “reasonable” assumption might be, of how a “reasonable” (and much lower) assumption might impact its projections for US household formations (possibly because doing so involves a lot of work.)