Behind the Eight Ball: Housing and Population Growth

News articles about U.S. housing frequently posit a shortfall in the supply of new housing relative to demand. E.g., the lead paragraph of a March 18 Wall Street Journal (WSJ) article reads, “America is facing a new housing crisis. A decade after an epic construction binge, fewer homes are being built per household than at almost any time in U.S. history.” But is that true; is the population-adjusted rate of new home construction really near an historical low? After all, if new supply is keeping pace with population growth, is there really a crisis?

To begin exploring this topic, we derived a metric -- called “starts per person added” (SPPA) -- in which monthly not-seasonally adjusted single- and multi-family starts are divided by the monthly change in the U.S. population. As Figures 1 and 2 below indicate, the cyclicality of population changes and construction activity within and among calendar years creates considerable month-over-month and year-over-year variability in both the single-family and multi-family metrics (Figure 3). Comparing long-term averages (Table 1) against recent results – and especially 12-month moving averages (Figure 4), which reduce seasonality effects and make broader trends more apparent -- reveals interesting patterns, however.

Figure 1. U.S. population and population growth, by month, since 1960. Source: U.S. Dept. of Commerce, Bureau of Economic Analysis (BEA). Click image for larger view.

Figure 2. Not-seasonally adjusted monthly single- and multi-family housing starts since 1960. Source: U.S. Census Bureau (CB). Click image for larger view.

Figure 3. Monthly single- and multi-family housing starts per person added to the population, since 1960. Sources: BEA, CB and Delphi Advisors. Click image for larger view.

For the entire period of January 1959 to May 2018 the average single-family SPPA = 0.405 (Table 1); the pre-2000 average = 0.420 (i.e., 0.420 single-family start per person added to the population). Put another way, prior to 2000, one single-family home was started for every 2.38 people added to the population. We selected 2000:01 as a breakpoint because it is a convenient middle ground among opinions regarding when the housing “bubble” began. During 2000:01-2007:07, the period encompassing the housing “bubble” and early part of the subsequent crash, the single-family SPPA averaged 0.520 -- reflecting the overwhelming market preference for single-family units.

Table 1. Average monthly housing starts per person added to the population during selected time intervals, by type of housing. Sources: BEA, CB and Delphi Advisors. Click image for larger view. 

The 1959:01-2018:05 average multi-family SPPA = 0.174; the pre-2000 average = 0.198. I.e., pre-2000, one multi-family unit was started for every 5.05 people added to the population. During 2000:01-2007:07 the multi-family SPPA averaged 0.123, again confirming the preference for single-family units.

As mentioned above, we calculated 12-month moving averages of the monthly data to reduce seasonality effects and to make broader trends easier to see (Figure 4). In the following discussion, we refer to the 12-month moving averages of single-family SPPA as “SPPA1F” and “SPPAMF” for multi-family.

Figure 4. Twelve-month moving average, or “MA(12),” of monthly single- (SPPA1F) and multi-family (SPPAMF) housing starts per person added to the population, since 1960. Sources: BEA, CB and Delphi Advisors. Click image for larger view.

Between 2000:01 and 2007:07, SPPA1F exceeded its pre-2000 average during 84 of those 91 months, peaked at 0.628 in November 2005, and averaged 0.520. Although SPPA1F has been consistently rising from its October 2009 minimum of 0.162 (2009:10-present average = 0.277), it has yet to regain the pre-2000 average. When adjusted for population growth, the single-family housing bust (which helped spawn the Great Recession) and its aftermath was deeper and, to date, has been over 50% longer than the preceding bubble.

The multi-family segment exhibits a different pattern. SPPAMF rebounded off its October 2009 low of 0.021 (2009:10-present average = 0.127), but has been gradually subsiding on trend since February 2017’s peak of 0.174. Like its single-family counterpart, SPPAMF has yet to regain its pre-2000 average.

These two metrics would suggest, then, that housing supply has indeed failed to keep pace with population growth. Moreover, unless multi-family starts again trend higher, the gap could continue widening. However, our findings do not confirm the WSJ’s contention that starts are presently near an all-time low.

The above paragraph led us to wonder how many more/fewer units have been built over time than might be justified by population growth. Attempting to answer that question requires a review of Figure 1. Notice that prior to 1990, monthly population growth was bouncing around in a range between roughly 150,000 and 225,000. Beginning in 2Q1990, however, population growth rocketed higher such that 225,000 became a “floor” rather than a “ceiling;” moreover, that higher trend growth rate was sustained through the subsequent decade. The Census Bureau observes “the population growth of 32.7 million people between 1990 and 2000 represents the largest census-to-census increase in American history,” but cautions “this increase may be caused by changes in census coverage, as well as births, deaths, and net immigration.” I.e., the jump may have been as much an artifact of statistical methodology as genuine, organic growth.

The reason for making the above observation is that SPPA1F and SPPAMF both tumbled below their respective long-term averages during the 1990s at least in part because estimates of population growth “reset” to a much higher pace. To account for the population-growth “step function” in 1990 when estimating what housing starts might be justified by population growth, we regrouped the SPPA data into two new intervals: pre-1990 and 1990:01-present. The bottom two rows of Table 1 present the relevant SPPA averages. If, for example, between 1959:01 and 1989:12 SPPA1F fell below 0.445 during a particular month, fewer single-family houses were built than justified by population growth. The “shortfall” in starts can be estimated by multiplying that month’s deviation from 0.445 by the corresponding population growth for that month. For comparisons starting in 1990, the average single-family SPPA was reset to 0.359 as a reflection of the population-growth step function. December 2017 was chosen as the ending point for estimating the SPPA averages since that date is now outside the range of data subject to monthly Census Bureau revisions.

Monthly deviations were accumulated over time; both the monthly and cumulative deviations for single-family starts are presented in Figure 5. It is worth noting that, although monthly deviations have been both positive and negative over time, the cumulative deviation remained negative (bottoming at -2.142 million units in October 1971) until August 2003. Consistent positive monthly deviations after February 1998 ultimately drove the cumulative deviation to a peak 2.348 million units in December 2007. Negative monthly deviations between 2008:01 and 2017:07 subsequently erased the “excesses” of the housing bubble. Although the cumulative deviation line has been turning upward since August 2017, it remains negative; as of May 2018, the cumulative deviation stood at -384,000 units. 

Figure 5. Monthly and cumulative deviations of single-family housing starts relative to interval averages. Sources: BEA, CB and Delphi Advisors. Click image for larger view.

Monthly deviations in the multi-family component have also oscillated both positive and negative over time (Figure 6). Interestingly, the cumulative deviation made it into positive territory during one 23-month period (1974:01 to 1975:10) -- the result of a flood of multi-family construction after 1967:01, perhaps related to Great Society programs. Since then, however, the cumulative deviation has remained negative; the near-term low was -1.042 million units in 4Q2012. The rebound in multi-family starts since 2009:10 has helped to trim the negative cumulative deviation, but as of 2018:05 it stood at -447,000 units.

Figure 6. Monthly and cumulative deviations of multi-family housing starts relative to interval averages. Sources: BEA, CB and Delphi Advisors. Click image for larger view.

To summarize, our analysis -- whether in terms of starts per person added to the population, or cumulative deviations in the number of units started relative to the number that might be justified by population growth -- shows that residential construction has not kept up with population changes. Perhaps this is a piece of the puzzle explaining why purchase- and rental-price affordability continue to be issues in many parts of the country. Particularly for the multi-family market, our findings also highlight potential upside for mass-timber market applications.

This analysis also buttresses our perception that the Millennial demographic wave has yet to truly enter the “shelter” market as its predecessors did at comparable stages of life. High shelter costs and other financialhurdles being experienced by the Millennial cohort likely mean more people per housing unit (both single- and multi-family) going forward. Finally, it implies that any Millennial demographic-led housing-start peak may not only be delayed but also more muted than otherwise expected as a function of the cohort’s size.