Blog Post < Previous | Next >
Guy
Generating a realistic California population projection
A couple of weeks ago I posted an essay on generating a realistic Marin County population projection. I uncovered that Sacramento's population projections for Marin County were way too high because of their overly aggressive migration assumptions. I wanted to check if this was the case for California as a whole.
My data sources for historical and projected population growth are the same as I used in my earlier essay:
a) For California, I use the California Department of Finance Demographic Research Unit (DRU);
b) For the US, I use the UN Population Division as disclosed on the website Our World in Data.
First some background on California's population trajectory
California's population trajectory can be divided into three periods:
1) The Boom years from 1950 to 1989. Over this period, California grew much faster than the US;
2) Convergence from 1990 to 2004. Over this period, California's population growth converged with the US.
3) The Down years from 2005 to the present when California's population grew much slower than the US.
You can see these three different periods visual trends in the graphs below.
The table below quantifies the visual data shown above.
California's population growth has drastically slowed down since 2005. Keep in mind the above data does not capture the COVID & Work From Home (WFH) period. As we will see WFH has caused California's population to shrink rapidly.
A review of the DRU California demographic projections
The graph below shows the DRU projection of California's natural growth rate ((births - deaths)/population) and migration. All the figures are shown in % change per year.
Within the graph above, the blue line denoting California's natural growth rate shows the impact of continuing demographic aging. On a stand-alone basis, this natural growth rate forecast is reasonable.
The DRU migration forecast (red line) is problematic. Remember this DRU forecast does not capture the WFH phenomenon. After peaking in 2011 at 0.29% p.a., migration dropped very rapidly and went negative to close to - 0.40% in 2019. Yet, the DRU forecasts that migration would quickly rebound to 0.31% and exceeds its former peak in 2011 (0.29%), and remain near a record level till the end of the forecast (2060).
Next, let's compare the DRU California projections vs. the UN US Medium scenario.
The graph below confirms again that the DRU California natural growth projections seem reasonable as they come in a bit lower than for the US. That is because California is expected to age a bit faster than the US.
When looking at migration (graph below), it shows that the DRU California projections pretty much match or exceed the UN US Medium scenario. This is highly unrealistic because the US migration rate is far higher than California's during the data history (2009 - 2019).
Due to WFH, the DRU projections are obsolete
The table below compares the actual data from 2020 to 2023 vs. the DRU projections.
As shown within the table above, the DRU California projections overestimate California's population by over 1.4 million individuals by 2023!
The main cause behind the DRU's population overestimation is its migration overestimation as shown in the table below.
The Actual est. figures entail some arithmetic estimation of what the migration rate should have been during this period given the resulting actual population figures and the DRU forecast of the natural growth rate (that is reasonable).
The DRU expected that migration would quickly rebound after 2019. Instead, because of WFH it actually cratered to record negative levels.
Comparing DRU projections vs. others out to 2060
I compared DRU projections vs. several others, including:
a) My own that I call California 1. I constructed my forecast:
i) using actual population data up to 2023;
ii) using the DRU natural growth rate projections;
iii) assuming migration would rebound to 0% by 2026 and to 0.145% by 2028. The 0.145% migration rate is at the 80th percentile of the migration rate history over the 2009 to 2023 period.
b) Two UN population projections for the US, using their US Low and US Medium scenarios.
The facet graph below shows all 4 projections.
The California DRU scenario as shown above looks way too high because:
- It totally misses the WFH impact over the 2020 - 2023 period as shown in my California 1 projection. Remember the 2020 - 2023 data points within my projections are not forecasted; they are actuals.
- It aligns way too long with the US Medium scenario when factoring that California has grown much slower than the US since 2005. Also, California has felt a severe impact from WFH that does not affect the US since it is not affected by population movements between States.
Comparing DRU projections vs. others out to 2030
Focusing on a shorter time horizon, out to 2030, renders the DRU projections even more unrealistic.
As shown in the facet graph above, my California 1 scenario that captures actual data through 2023 suggests that California's population may come close to recovering its former 2020 level by 2030. The California DRU scenario instead starts at the 2020 level, ignores the cratering over the 2020 - 2023 period, and goes straight up from there projecting that the California population will be 5% higher in 2030 vs. 2020. As depicted, the California DRU projections are already obsolete.
Documenting the DRU overstating migration rate
The table above shows the distribution of migration rates within the historical data (2009 - 2023) vs. the DRU projections out to 2030 and 2060. It shows both the maximum and minimum and numerous percentiles in between.
The DRU projected migration rates are way too high. You can observe that in several ways.
- The DRU projections for the 75th percentile at 0.30% are not even in the historical data, as they exceed the historical maximum of 0.29%.
- The DRU projections out to 2030 have a 5th percentile of 0.08% which is twice as high as the Median (50th percentile) for history. The DRU out to 2060 has a 5th percentile at 0.17% which is much higher than the 75th percentile in history at 0.14%. It is also higher than the 80th percentile I used for my projection at 0.145%.
Conclusion
The DRU California population projections are way too high. By 2023, they already exceed actual population figures by over 1.4 million. The DRU needs to revise such projections asap.