BELLINGHAM — In mid-September, the U.S. Census Bureau released the results of its annual American Community Survey. The report, based on nearly 2 million responses, contains nationwide 2023 population, demographic and economic data that can be broken down by state, county and city.
Their estimates for Bellingham were concerning. The data suggested a $12,000 drop in median household income — from $67,419 in 2022 to $54,867 last year — and a 54% increase in poverty. Local experts, though, have their doubts.
“The first thing to remember about the ACS data is that when you get to a small area, that rule [is] any one year’s figure is not going to be that precise,” Hart Hodges, an economics professor at Western Washington University, told the Bellingham Herald.
Changes to Census Bureau methodology
However, that doesn’t tell the whole story. One-year numbers have remained relatively steady in the past — Bellingham’s median household income was between $65,000 and $70,000 each year between 2019 and 2022. The Census Bureau accounts for the relatively small size of the sample with its margin of error, which it set at $4,075, meaning there’s a 90% chance that the true median is within that amount of their estimate if they used a proper sample.
“There were no methodological changes to income in the ACS between 2022 and 2023,” said Kirby Posey, a survey statistician with the Census Bureau. “The difference between the medians is still statistically significant, but we should consider the margins of error when considering the magnitude of the change.”
Hodges suggests that a change in methodology prompted by low response-rates COVID-19 pandemic is to blame.
“There’s an issue that’s not getting a lot of attention, especially in the public press, because it’s a bit wonky,” Hodges said. “Prior to COVID, when the census or Bureau of Labor Statistics or somebody else is doing surveys … response rates were above 50%, which meant, if I’m contacting this many households and over half of them were responding, I’m getting a good mix.”
With lower response rates, even if officials sent surveys to a sample that represented each demographic in the city proportionately, they could no longer guarantee that the end result would properly represent each group.
“In COVID, response rates started to drop,” Hodges said. “For a lot of these surveys, response rates fell to 30%, and now you have to say, if I survey 100 households and only 30 reply, I may not have a good mix. It might be 30% spread evenly over ethnic groups or over income. But it might not.”
On the ACS, the overall response rate fell from 62% in the five years prior to the pandemic, to between 55% and 56% in each of the three years since. Bellingham’s response rate saw an even more significant change, going from 74% from 2015 to 2019, to 67% and 68% the past two years.
“This has been a challenging period for data collection,” Erica Gardner, a research scientist at the Washington State Office of Financial Management, said in an email to the Bellingham Herald. “The 2020 ACS was not released due to the disruption in data collection caused by COVID pandemic. Response rates to surveys took a hit after the pandemic and have yet to fully recover.”
In order to account for the change, the Census Bureau began putting more weight into data from groups that tended to respond at a lower rate, according to Hodges and Gardner.
“They’re starting to say, ‘Oh, wait, we’re getting higher response rates from people with these characteristics.’ And they can either go back out and survey again, or they can start weighting the responses,” Hodges said.
However, that can lead to problems, especially with smaller data sets. If, for one year in one city, a group that tends to respond at a low rate responds at the same rate as everyone else, their numbers will be over-represented in citywide data.
“I’ll make this up to illustrate. If I’m getting fewer responses from low income households, and the historical average response rates for high income households, maybe I’ll double count the one side,” Hodges said. “What if you do that in 2022, and then in 2023 you get a historical average [response rate], but you’re weighting it?”
Hodges suspects that this could be part of the reason Bellingham’s income numbers fell so sharply.
“It’s 477 households in the survey, right?” Hodges said. “All it would have taken was 20 or 30 or 40 low-income households responding who hadn’t responded in the prior two years.”
Bellingham’s student population affects data
The chances of that sort of mistake occurring are even higher in college towns, according to both Hodges and Gardner. Western Washington University’s nearly 15,000 account for roughly 15% of the city’s population, which likely contributes to Bellingham’s high poverty rate as well.
“While Bellingham is a larger city with its own characteristics, the college population does add in another dimension of uncertainty,” Gardner said. “College towns typically have higher poverty rates due to the student population, but the higher rates tend to be concentrated in those student ages.”
What does WA data say?
State data support Hodges’ theory. The Washington State Employment Security Department puts together annual estimates of the average annual salary for employees in each city and county in the state. While the ESD hasn’t finalized its 2023 estimates for Whatcom County, according to Jim Vleming, the department’s regional labor economist for northwest Washington, the preliminary numbers actually show a rise in wages.
In an email to the Herald, Fleming said that the department’s preliminary estimate of the 2023 average wage in Whatcom County is $61,900, up from $59,174 the previous year and $56,216 in 2021.
“This wage data is collected by employers, covering all employers in the county that have tax accounts with the state,” Vleming said. “So this is a pretty complete picture of what is happening in Whatcom County.”
If wages in Whatcom County didn’t fall in 2023, Hodges says it’s unlikely that Bellingham, which makes up roughly 40% of the county, saw a sharp decrease in its average income.
“It’s hard to imagine that there would be a big decrease in wages in the city when it’s not visible in the county,” Hodges said.
If there really was a loss in income that didn’t show up in the local wage data, Hodges said it would likely have to be caused by a business closure.
“So then the next layer is, were there business closures in the city?” Hodges said. “If you pull Faith Life out, or if you pull the whole cardiology department out of PeaceHealth — what would it take to see that big of a drop?”
Gardner said that if citywide income levels did fall by 12%, it would likely be apparent to experts across the city.
“If there is nothing that local experts can point to that would explain the decline, there likely wasn’t a decline or at least a decline as pronounced as shown in this year’s data,” Gardner said.
While there were a few notable business closures in Bellingham that year, Hodges said that none of them were significant enough to lead to such a sharp drop off in income. It’s still possible that Bellingham’s average income fell last year, but Hodges doubts that the drop was nearly as significant as the Census Bureau estimates.
“So I just couldn’t find other data that said there’s that big of a change, somehow, somewhere, to make me think that estimate is plausible,” Hodges said. “Everything I was looking at was pointing to ‘I bet that estimate is an outlier.’”
