Is the Region Growing Fast or Slow?
Last month we asked our readers what they thought about discordant data on growth in the greater Madison region. The poll was prompted by an InBusiness article that cited some surprising data first reported in the Wisconsin State Journal. Dane County, it would seem, had grown by only 5,600 people in three years.[1]
In the poll, we asked you to choose between US Census Bureau data indicating slower growth or Wisconsin Department of Administration (DOA) population estimates indicating more rapid growth. Around 90% of respondents indicated that they agreed with DOA that population growth in Dane County has progressed more rapidly since 2020.
So which data are right? The answer—like so many answers surrounding data—is not a simple one.
The Census Bureau currently reports a population increase of 5,600 residents from July 2020 through July 2022 in the annual Population Estimates Program (PEP) estimates. DOA, by contrast, reports a population change of almost 20,700 between April 2020[2] and January 2022, a monthly growth rate almost three times the one estimated by the Census Bureau!
How is it possible that these two estimates are so different? Is there a difference in how the Census Bureau and DOA collect their data, define “total population,” or calculate change? No, not a significant one. Population change is calculated in virtually the same way at the DOA and Census Bureau. Cohort change is calculated using a common standard: births minus deaths plus net migration. As for the data sources, most are comparable or identical.[3] So, what gives?
A portion of the answer can be found in how the Census Bureau employs control totals for its population and household estimates.[4] All other data products, like the American Community Survey (ACS), rely on these estimates. The Census Bureau ensures that state estimates are controlled to a national control total, counties are controlled to the state, and so on. The methodology is built on the premise that estimating population totals is easier and more accurate on a large scale.
DOA takes a slightly different approach with county and state control totals, using them for only some of the projections.[5] The primary difference between the two agencies’ estimates is top-down versus bottom-up estimation. But this only accounts for some of the discrepancy.
Signs also point to the effects of the COVID-19 Pandemic being the culprit. Media coverage and delayed release dates for 2020 Census data all underscored the difficulties experienced in the 2020 enumeration.[6] But post-survey analysis of the 2020 Census suggests that coverage errors were fewer than in 2010. The Post Enumeration Survey (PES) found Wisconsin’s rate of under-/over-count not statistically different from zero. The Demographic Analysis (DA) process further corroborated this.
However, the data look different when you begin to analyze finer grain details of age groups. The DA results show that the population aged 18-24 was overcounted nationwide. The PES indicates an undercount of persons aged 18-29, which might seem slightly at odds with the preceding sentence.[7] The key factor is that the PES does not count the population in group quarters, i.e., dorms, but the DA does.
| NET MIGRATION | |||||
| Calendar Year | 2017 | 2018 | 2019 | 2020 | 2021 |
| Dane County | 19,487 | 23,994 | 27,603 | 34,507 | 17,486 |
| Wisconsin | (1,825) | 14,921 | 15,952 | 64,910 | 56,928 |
In mid-March of 2020, college students across the state and nation were packing up their dorm rooms and returning home. Undercounting college students where you would expect them—in UW Madison residence halls—and overcounting them in other places outside the county or the state, could be the remaining puzzle piece in the divergent population growth estimates.
Interestingly, net migration in Dane County did jump in 2020 and in Wisconsin during 2020 and 2021 (See Table). Combine this occurrence with how and at what geographic level the DOA and Census Bureau use population control totals to generate their estimates, and we have a working theory for why one source lists a growth rate many times larger than the other.
Unfortunately, exactly why these data discrepancies exist must remain a theory for now. The Census Bureau is currently looking into how student migration during the 2020 Census may have impacted coverage error and enumeration.
But take heart, settling the score on our poll and recommending which source of population data to believe is an easier question to answer. The Census Bureau produces the authoritative population count data. Numerous funding sources and the apportionment of the US House of Representatives requires their use. When citing the number/percentage of people for descriptive statistics, use the Decennial Census or other authorized Census Bureau product. When citing population and household growth rates, we recommend the DOA’s estimates at present. There are numerous, trustworthy datasets that can confirm housing and population growth continue at a rapid pace.
DOA will release preliminary population estimates for January 1, 2023 by August 10 and final population estimates by October 10. We will report back in a future newsletter on the new population estimates and additional Census Bureau reporting.
[1] The InBusiness article incorrectly states the period of change as three years. Data referenced in the articles were taken from the US Census Bureau’s Population Estimates Program (PEP). PEP products are updated in July of each year. Current vintages of the PEP supersede all previous data.
[2] DOA lists Decennial Census data as official totals, rather than their own estimates from January 1st of the corresponding year.
[3] This may be enough of an answer for some. Comparing different datasets with even slightly different methodologies can be fraught with problems.
[4] Details can be found here.
[5] https://doa.wi.gov/DIR/PopEstMethod.pdf
[6] Survey non-response for the concurrent American Community Survey in 2020 was such an issue that the Census Bureau declined to release an official 1-year estimate. It is possible that 2021 product may also exhibit irregularities due to the pandemic.
[7] There are, unfortunately, no smaller age categories from the PES that would pinpoint the issue to the 18-24 age group. It is possible, but unlikely, that the population aged 25-29 was massively undercounted, counteracting an overcount of 18-24 year olds.