The Marin Post

The Voice of the Community

Blog Post < Previous | Next >

Guy

Sacramento has a long history of overestimating population growth

Let's evaluate how well the Department of Finance (DoF) has done in forecasting California's population in the past. I will use a method of testing their model that is called either backtesting, out-of-sample testing or Hold Out testing. All these methods amount to doing the same thing. And, that is to look at a model's earlier forecasts and see how well they fit the actual figures as time goes by.

From conducting such tests, you can expect one of three outcomes:

  1. The model errors are random. Some of the errors are overestimated, and others are underestimated. This reflects an objective model without explicit bias;
  2. The model errors consistently overestimate the population. In this case, you have identified an upward bias within the model methodology;
  3. The model errors consistently underestimate the population. Now, this would reflect a downward bias within the model methodology.

The challenge is how to identify the DoF earlier model version forecasts. I did not find any archived information on their website. However, instead, I did find some estimates from the relevant Media.

Dan Walters' article California's shrinking population at CalMatters published on April 10, 2022, was a great source of information. Quoting Dan Walters below:

“Although California’s population growth began to slow in the 1990s after exploding in the previous decade by 6 million people, both official and independent demographers continued to see relatively strong growth for decades to come.

In 2007, then-Gov. Arnold Schwarzenegger’s in-house demographers projected that California would have 39.9 million residents by 2011. It didn’t happen.

Five years later, then-Gov. Jerry Brown’s 2012-13 budget projected that the state’s population would be “over 39.6 million” by 2016. That didn’t happen either.”

Ethan Varian's article California population projected to stagnate through 2060 published by the Bay Area Newsgroup on July 26, 2023, was also informative. It discloses this spectacular graph showing how far off DoF's population projection was only 10 years ago back in 2013.

2013-forecast.png

As shown above, only 10 years ago the DoF projected that California's population would reach 52.7 million by 2060. Meanwhile, in its most current forecast, the DoF now projects that California's population will remain pretty much flat at 39.5 million.

Extracting all the text and visual information I could from the two mentioned articles, I back-tested the DoF models going back to 2007.

Backtesting.png

Click on the image to enlarge

The numbers are a bit small. So, let me walk you through the first row of this table from left to right.

So, back in 2007, DoF projected that the California population would grow from 36.55 million (in 2007) to 39.9 million in 2011. Instead, the population increased only to 37.08 million. They overestimated population growth by 2.82 million. The projected growth was 9.2% instead of just 1.5%. The excess growth was 7.7% over just a 4-year period. The excess growth per year was 1.9%. This forecast based on demographic standards is way off just four years out into the forecast. Granted, the 2007 forecast is the worst when using the "excess growth per year" parameter that allows us to compare all the forecasts on an equal footing.

However, notice that the DoF forecasts made in July 2021 are nearly as bad. Their projections for the California population in 2021, 2022, and 2023 all have excess growth per year ranging from 1.2% to 1.8%. The population in 2020 was 39.53 million. And, DoF predicted population would reach 40.35 million in 2023. Instead, the population declined to 38.99 million. That is an error of nearly 1.4 million over just a 3-year period.

The DoF made its July 2021 forecast when we were already nearly a year and a half into the COVID and Work From Home (WFH) era. It was well known that individuals and even businesses were leaving the State in droves. Meanwhile, the DoF was projecting rapid migration into the State during this same period, anticipating that the population would grow rapidly.

From the above model back-testing exercise, we can readily conclude that the DoF forecasting methodology has a very strong upward bias.

It is interesting to observe the huge downward revision of the DoF long-term forecast out to 2060.

DoF-2060.png

The above denotes embarrassingly large revisions.

The bias outlook for the DoF forecast released in July 2023

After confirming that all the prior DoF forecasts have a very strong upward bias, can we anticipate if the most recent DoF forecast released in July 2023 will also have an upward bias?

Yes, we can. One of the DoF's methodology flaws is to chronically overestimate migration into the State. This is especially true at a time when migration flows have changed direction due to WFH and other numerous regulatory and economic impediments to conducting business in the State.

To benchmark the DoF migration rates, we will compare them with the ones for the US as projected by the UN Population Division (UN).

The table below describes the distribution of several migration rates from left to right:

  1. California History represents the migration rate during the history of the data from 2009 to 2022. (2020 – 2022 are based on solid estimates of births, deaths, and migrants).
  2. California Forecast represents the migration rates within the DoF forecast from 2023 to 2060.
  3. US History represents the US migration rates during the history of the data.
  4. US Forecast represents the US migration rates within the UN World Population Prospects forecast out to 2060.

migration-prospect.pngThe yellow-highlighted row is the 50th percentile, also called the Median

There are several takeaways from the table above:


Median-migration.png

California’s prospective migration rates are unlikely to increase over the historical ones for several reasons.

  1. The migration – feeders (other States, and other countries) are experiencing their own demographic aging. Thus, they will send fewer migrants to California;
  2. Deglobalization, trade wars, etc. will result in fewer migrants (outside of North America) having access to California;
  3. Work From Home (WFH) is and will have a permanent effect. Employees do not need to move to California cities anymore to have access to our world-class companies;
  4. California's rising costs at both the individual and business levels including taxes, utilities, insurance rates, etc.;
  5. California’s constraints: water scarcity, electricity grid fragility, traffic congestion & public transportation inadequacies.

Small differences in migration rates have a large impact on long-term projections. By just using the historical median migration rate of 0.05% instead of the DoF median during the forecast of 0.14%, the California population by 2060 is projected to be 38.2 million instead of 39.5 million. That is a material difference of minus 1.3 million or – 3.3%. It also changes the overall message. California’s population would be projected to explicitly decline a bit instead of just remaining stable.

Migration-two-graphs.pngThe prospect of California’s population declining a bit over the long term is not controversial once you become aware that this is a common trend throughout much of the developed World. Between 2020 and 2060, Europe’s population is projected to decline by – 9.4%, China by – 14.8% according to the UN Medium fertility scenario.

China-and-Europe.png

The two above examples are just representative of two huge population regions. But, they are not unique. There are just two regions or countries among many others with a common prospect of population decline due to demographic aging (readily apparent in the two graphs). Such countries with a similar fate include: Japan, South Korea, and Russia just to name a few (as mentioned just about all of Europe also faces population decline).

Relative to the mentioned regions or countries, California bears a substantial demographic handicap. And, that is it is very strongly impacted by the mentioned WFH effect.

THE END